From jonathan.aldrich at cs.cmu.edu Sun Jan 5 17:06:05 2020 From: jonathan.aldrich at cs.cmu.edu (Jonathan Aldrich) Date: Sun, 5 Jan 2020 17:06:05 -0500 Subject: [TYPES] Open Access petition to the ACM - please sign! Message-ID: All, Robert Rand, after conversations with a number of us in a Facebook group, has put together a petition to the ACM about open access. It's clear that many people on Types are passionate about this...if you are an ACM member, I encourage you to join me in signing it! https://www.change.org/p/association-for-computing-machinery-acm-support-open-access The petition asks ACM to remove its signature from the letter that opposed the proposed US open access policy, and asks for a vote of the membership on making the ACM digital library all open access, with any fees based on publication cost only. If 1% of ACM members sign, then ACM's constitution (article 6) says that they must put the text in italics to a vote of the membership. The full text of the petition is below. Cheers, Jonathan As members of the Association for Computing Machinery (ACM) we were disappointed to see the ACM sign onto a letter opposing a US government policy to require the free distribution of all federally funded research. As computer scientists, we are committed to the principles of Open Source, which promotes universal access to software, and Open Science, in which scientific research is both transparent and publicly accessible. When research funded by public agencies is signed away to private organizations and embargoed by them, it makes a mockery of these principles. As such, we demand that the ACM immediately withdraw its signature from the letter to the US government. Moreover, we ask that the following proposal be put to a vote of the general ACM membership: *The Association for Computing Machinery shall release all papers in the ACM digital library, past and future, under fully open access terms within five years. Moreover, it shall ensure that any charges for publication, whether to authors, conference attendees or institutions, should be no more than the cost of publishing the papers themselves.* We further encourage the leadership of the ACM to take concrete steps towards Open Access immediately, in line with its stated purpose of fostering the open interchange of information. From roberto at dicosmo.org Mon Jan 6 15:29:07 2020 From: roberto at dicosmo.org (Roberto Di Cosmo) Date: Mon, 6 Jan 2020 21:29:07 +0100 Subject: [TYPES] Open Access petition to the ACM - please sign! In-Reply-To: References: Message-ID: <20200106202907.GA15816@traveler> Dear Joathan, thanks for sharing this: it is an important step forward to make our voices heard, and I've been delighted to sign it! We need to take quite a few steps, though, on the still long path to set up a sustainable infrastructure for the research publishing workflow that can last on the long term, and avoid the pitfalls we have often fall into when engaging in Open Source (see https://www.fordfoundation.org/work/learning/research-reports/roads-and-bridges-the-unseen-labor-behind-our-digital-infrastructure/ for a "nice" museum of horrors :-)) I'll try to share some thoughts on this in the coming days Cheers -- Roberto On Sun, Jan 05, 2020 at 05:06:05PM -0500, Jonathan Aldrich wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > All, > > Robert Rand, after conversations with a number of us in a Facebook group, > has put together a petition to the ACM about open access. It's clear that > many people on Types are passionate about this...if you are an ACM member, > I encourage you to join me in signing it! > > https://www.change.org/p/association-for-computing-machinery-acm-support-open-access > > > The petition asks ACM to remove its signature from the letter that opposed > the proposed US open access policy, and asks for a vote of the membership > on making the ACM digital library all open access, with any fees based on > publication cost only. If 1% of ACM members sign, then ACM's constitution > (article 6) says that they must put the text in italics to a vote of the > membership. > > The full text of the petition is below. > > Cheers, > > Jonathan > > > As members of the Association for Computing Machinery (ACM) we were > disappointed to see the ACM sign onto a letter opposing a US government > policy to require the free distribution of all federally funded research. > As computer scientists, we are committed to the principles of Open Source, > which promotes universal access to software, and Open Science, in which > scientific research is both transparent and publicly accessible. When > research funded by public agencies is signed away to private organizations > and embargoed by them, it makes a mockery of these principles. > > As such, we demand that the ACM immediately withdraw its signature from the > letter to the US government. Moreover, we ask that the following proposal > be put to a vote of the general ACM membership: > > *The Association for Computing Machinery shall release all papers in the > ACM digital library, past and future, under fully open access terms within > five years. Moreover, it shall ensure that any charges for publication, > whether to authors, conference attendees or institutions, should be no more > than the cost of publishing the papers themselves.* > > We further encourage the leadership of the ACM to take concrete steps > towards Open Access immediately, in line with its stated purpose of > fostering the open interchange of information. -- Roberto Di Cosmo ------------------------------------------------------------------ Computer Science Professor (on leave at INRIA from IRIF/University Paris Diderot) Director Software Heritage https://www.softwareheritage.org INRIA Bureau C328 E-mail : roberto at dicosmo.org 2, Rue Simone Iff Web page : http://www.dicosmo.org CS 42112 Twitter : http://twitter.com/rdicosmo 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 ------------------------------------------------------------------ GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 From selinger at mathstat.dal.ca Wed Jan 8 12:53:38 2020 From: selinger at mathstat.dal.ca (selinger at mathstat.dal.ca) Date: Wed, 8 Jan 2020 17:53:38 +0000 Subject: [TYPES] Open Access petition to the ACM - please sign! In-Reply-To: <20200106202907.GA15816@traveler> Message-ID: Dear Roberto, thanks for this reference. I agree with your proposition that some work is needed to ensure that open access thrives in the long run. In fairness, though, I don't think the issues of open access publishing and open source software are quite analogous. The economic difference between open source software and proprietary software, and the source of the problems this article points out, is that people developing open source software typically volunteer their time, whereas people developing proprietary software are typically paid. In the publishing world, there is no such difference. The editors and referees of open access journals are unpaid; the editors and referees of closed access journals are also unpaid (except in some cases for a small pro-forma honorarium to the editor in chief). The extra revenue the ACM and other publishers are collecting does not go towards supporting the people who do most of the work of publishing. Neither does it typically go towards funding the underlying research. I signed the petition. Since the ACM has less than 100,000 members, only 1,000 signatures are required to trigger a vote. Best, -- Peter Roberto Di Cosmo wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Joathan, > thanks for sharing this: it is an important step forward to make our voices heard, and I've been delighted to sign it! > > We need to take quite a few steps, though, on the still long path to set up a sustainable infrastructure for the research publishing workflow that can last on the long term, and avoid the pitfalls we have often fall into when engaging in Open Source (see https://www.fordfoundation.org/work/learning/research-reports/roads-and-bridges-the-unseen-labor-behind-our-digital-infrastructure/ for a "nice" museum of horrors :-)) > > I'll try to share some thoughts on this in the coming days > > Cheers > > -- > Roberto > > On Sun, Jan 05, 2020 at 05:06:05PM -0500, Jonathan Aldrich wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > All, > > > > Robert Rand, after conversations with a number of us in a Facebook group, > > has put together a petition to the ACM about open access. It's clear that > > many people on Types are passionate about this...if you are an ACM member, > > I encourage you to join me in signing it! > > > > https://www.change.org/p/association-for-computing-machinery-acm-support-open-access > > > > > > The petition asks ACM to remove its signature from the letter that opposed > > the proposed US open access policy, and asks for a vote of the membership > > on making the ACM digital library all open access, with any fees based on > > publication cost only. If 1% of ACM members sign, then ACM's constitution > > (article 6) says that they must put the text in italics to a vote of the > > membership. > > > > The full text of the petition is below. > > > > Cheers, > > > > Jonathan > > > > > > As members of the Association for Computing Machinery (ACM) we were > > disappointed to see the ACM sign onto a letter opposing a US government > > policy to require the free distribution of all federally funded research. > > As computer scientists, we are committed to the principles of Open Source, > > which promotes universal access to software, and Open Science, in which > > scientific research is both transparent and publicly accessible. When > > research funded by public agencies is signed away to private organizations > > and embargoed by them, it makes a mockery of these principles. > > > > As such, we demand that the ACM immediately withdraw its signature from the > > letter to the US government. Moreover, we ask that the following proposal > > be put to a vote of the general ACM membership: > > > > *The Association for Computing Machinery shall release all papers in the > > ACM digital library, past and future, under fully open access terms within > > five years. Moreover, it shall ensure that any charges for publication, > > whether to authors, conference attendees or institutions, should be no more > > than the cost of publishing the papers themselves.* > > > > We further encourage the leadership of the ACM to take concrete steps > > towards Open Access immediately, in line with its stated purpose of > > fostering the open interchange of information. > > -- > Roberto Di Cosmo > > ------------------------------------------------------------------ > Computer Science Professor > (on leave at INRIA from IRIF/University Paris Diderot) > > Director > Software Heritage https://www.softwareheritage.org > INRIA > Bureau C328 E-mail : roberto at dicosmo.org > 2, Rue Simone Iff Web page : http://www.dicosmo.org > CS 42112 Twitter : http://twitter.com/rdicosmo > 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 > ------------------------------------------------------------------ > GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 > From roberto at dicosmo.org Wed Jan 8 13:37:01 2020 From: roberto at dicosmo.org (Roberto Di Cosmo) Date: Wed, 8 Jan 2020 19:37:01 +0100 Subject: [TYPES] Open Access petition to the ACM - please sign! In-Reply-To: References: <20200106202907.GA15816@traveler> Message-ID: Hi Peter, thank you for signing: this kind of pressure is quite needed to move things forward :-) I plan to come back with a more detailed explanation of the pitfalls we need to avoid, which indeed do not emerge necessarily clearly from Nadia Eghbal's account: in a nutshell, when looking for long term sustainability, we need to make sure we take into account all the labour and costs that too often go unseen, and look for a solution that is properly architectured to produce economies of scale, and funnel the benefits back to stakeholders, not shareholders. In this respect, I feel the "transformative" plan for Open Access that ACM put forward almost the same day it signed the infamous letter, contains a potentially good idea that needs investigation (and of course, clearer/more transparent cost calculations :-)) -- Roberto ------------------------------------------------------------------ Computer Science Professor (on leave at INRIA from IRIF/University Paris Diderot) Director Software Heritage E-mail : roberto at dicosmo.org INRIA Web : http://www.dicosmo.org Bureau C328 Twitter : http://twitter.com/rdicosmo 2, Rue Simone Iff Tel : +33 1 80 49 44 42 CS 42112 75589 Paris Cedex 12 ------------------------------------------------------------------ GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 Le mer. 8 janv. 2020 ? 18:53, a ?crit : > Dear Roberto, > > thanks for this reference. I agree with your proposition that some > work is needed to ensure that open access thrives in the long run. In > fairness, though, I don't think the issues of open access publishing > and open source software are quite analogous. > > The economic difference between open source software and proprietary > software, and the source of the problems this article points out, is > that people developing open source software typically volunteer their > time, whereas people developing proprietary software are typically > paid. > > In the publishing world, there is no such difference. The editors and > referees of open access journals are unpaid; the editors and referees > of closed access journals are also unpaid (except in some cases for a > small pro-forma honorarium to the editor in chief). The extra revenue > the ACM and other publishers are collecting does not go towards > supporting the people who do most of the work of publishing. Neither > does it typically go towards funding the underlying research. > > I signed the petition. Since the ACM has less than 100,000 members, > only 1,000 signatures are required to trigger a vote. > > Best, -- Peter > > Roberto Di Cosmo wrote: > > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > Dear Joathan, > > thanks for sharing this: it is an important step forward to make > our voices heard, and I've been delighted to sign it! > > > > We need to take quite a few steps, though, on the still long path to set > up a sustainable infrastructure for the research publishing workflow that > can last on the long term, and avoid the pitfalls we have often fall into > when engaging in Open Source (see > https://www.fordfoundation.org/work/learning/research-reports/roads-and-bridges-the-unseen-labor-behind-our-digital-infrastructure/ > for a "nice" museum of horrors :-)) > > > > I'll try to share some thoughts on this in the coming days > > > > Cheers > > > > -- > > Roberto > > > > On Sun, Jan 05, 2020 at 05:06:05PM -0500, Jonathan Aldrich wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > > > All, > > > > > > Robert Rand, after conversations with a number of us in a Facebook > group, > > > has put together a petition to the ACM about open access. It's clear > that > > > many people on Types are passionate about this...if you are an ACM > member, > > > I encourage you to join me in signing it! > > > > > > > https://www.change.org/p/association-for-computing-machinery-acm-support-open-access > > > > > > > > > The petition asks ACM to remove its signature from the letter that > opposed > > > the proposed US open access policy, and asks for a vote of the > membership > > > on making the ACM digital library all open access, with any fees based > on > > > publication cost only. If 1% of ACM members sign, then ACM's > constitution > > > (article 6) says that they must put the text in italics to a vote of > the > > > membership. > > > > > > The full text of the petition is below. > > > > > > Cheers, > > > > > > Jonathan > > > > > > > > > As members of the Association for Computing Machinery (ACM) we were > > > disappointed to see the ACM sign onto a letter opposing a US government > > > policy to require the free distribution of all federally funded > research. > > > As computer scientists, we are committed to the principles of Open > Source, > > > which promotes universal access to software, and Open Science, in which > > > scientific research is both transparent and publicly accessible. When > > > research funded by public agencies is signed away to private > organizations > > > and embargoed by them, it makes a mockery of these principles. > > > > > > As such, we demand that the ACM immediately withdraw its signature > from the > > > letter to the US government. Moreover, we ask that the following > proposal > > > be put to a vote of the general ACM membership: > > > > > > *The Association for Computing Machinery shall release all papers in > the > > > ACM digital library, past and future, under fully open access terms > within > > > five years. Moreover, it shall ensure that any charges for publication, > > > whether to authors, conference attendees or institutions, should be no > more > > > than the cost of publishing the papers themselves.* > > > > > > We further encourage the leadership of the ACM to take concrete steps > > > towards Open Access immediately, in line with its stated purpose of > > > fostering the open interchange of information. > > > > -- > > Roberto Di Cosmo > > > > ------------------------------------------------------------------ > > Computer Science Professor > > (on leave at INRIA from IRIF/University Paris Diderot) > > > > Director > > Software Heritage https://www.softwareheritage.org > > INRIA > > Bureau C328 E-mail : roberto at dicosmo.org > > 2, Rue Simone Iff Web page : http://www.dicosmo.org > > CS 42112 Twitter : http://twitter.com/rdicosmo > > 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 > > ------------------------------------------------------------------ > > > GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 > > > > > From b.a.w.spitters at gmail.com Thu Jan 9 02:13:57 2020 From: b.a.w.spitters at gmail.com (Bas Spitters) Date: Thu, 9 Jan 2020 08:13:57 +0100 Subject: [TYPES] Open Access petition to the ACM - please sign! In-Reply-To: References: <20200106202907.GA15816@traveler> Message-ID: Let me add a European perspective to this. Most European funders will require OA publication starting 2021! https://www.coalition-s.org/ https://avandeursen.com/2019/08/20/europes-open-access-plan-s-and-paper-publishing-in-software-engineering-research/ The aim of these Plan S ?funders? (collectively called ?Coalition S?), is that "With effect from 2021, all scholarly publications on the results from research funded by public or private grants provided by national, regional and international research councils and funding bodies, must be published in Open Access Journals, on Open Access Platforms, or made immediately available through Open Access Repositories without embargo." On Thu, Jan 9, 2020 at 7:18 AM wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Roberto, > > thanks for this reference. I agree with your proposition that some > work is needed to ensure that open access thrives in the long run. In > fairness, though, I don't think the issues of open access publishing > and open source software are quite analogous. > > The economic difference between open source software and proprietary > software, and the source of the problems this article points out, is > that people developing open source software typically volunteer their > time, whereas people developing proprietary software are typically > paid. > > In the publishing world, there is no such difference. The editors and > referees of open access journals are unpaid; the editors and referees > of closed access journals are also unpaid (except in some cases for a > small pro-forma honorarium to the editor in chief). The extra revenue > the ACM and other publishers are collecting does not go towards > supporting the people who do most of the work of publishing. Neither > does it typically go towards funding the underlying research. > > I signed the petition. Since the ACM has less than 100,000 members, > only 1,000 signatures are required to trigger a vote. > > Best, -- Peter > > Roberto Di Cosmo wrote: > > > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > Dear Joathan, > > thanks for sharing this: it is an important step forward to make our voices heard, and I've been delighted to sign it! > > > > We need to take quite a few steps, though, on the still long path to set up a sustainable infrastructure for the research publishing workflow that can last on the long term, and avoid the pitfalls we have often fall into when engaging in Open Source (see https://www.fordfoundation.org/work/learning/research-reports/roads-and-bridges-the-unseen-labor-behind-our-digital-infrastructure/ for a "nice" museum of horrors :-)) > > > > I'll try to share some thoughts on this in the coming days > > > > Cheers > > > > -- > > Roberto > > > > On Sun, Jan 05, 2020 at 05:06:05PM -0500, Jonathan Aldrich wrote: > > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > > > All, > > > > > > Robert Rand, after conversations with a number of us in a Facebook group, > > > has put together a petition to the ACM about open access. It's clear that > > > many people on Types are passionate about this...if you are an ACM member, > > > I encourage you to join me in signing it! > > > > > > https://www.change.org/p/association-for-computing-machinery-acm-support-open-access > > > > > > > > > The petition asks ACM to remove its signature from the letter that opposed > > > the proposed US open access policy, and asks for a vote of the membership > > > on making the ACM digital library all open access, with any fees based on > > > publication cost only. If 1% of ACM members sign, then ACM's constitution > > > (article 6) says that they must put the text in italics to a vote of the > > > membership. > > > > > > The full text of the petition is below. > > > > > > Cheers, > > > > > > Jonathan > > > > > > > > > As members of the Association for Computing Machinery (ACM) we were > > > disappointed to see the ACM sign onto a letter opposing a US government > > > policy to require the free distribution of all federally funded research. > > > As computer scientists, we are committed to the principles of Open Source, > > > which promotes universal access to software, and Open Science, in which > > > scientific research is both transparent and publicly accessible. When > > > research funded by public agencies is signed away to private organizations > > > and embargoed by them, it makes a mockery of these principles. > > > > > > As such, we demand that the ACM immediately withdraw its signature from the > > > letter to the US government. Moreover, we ask that the following proposal > > > be put to a vote of the general ACM membership: > > > > > > *The Association for Computing Machinery shall release all papers in the > > > ACM digital library, past and future, under fully open access terms within > > > five years. Moreover, it shall ensure that any charges for publication, > > > whether to authors, conference attendees or institutions, should be no more > > > than the cost of publishing the papers themselves.* > > > > > > We further encourage the leadership of the ACM to take concrete steps > > > towards Open Access immediately, in line with its stated purpose of > > > fostering the open interchange of information. > > > > -- > > Roberto Di Cosmo > > > > ------------------------------------------------------------------ > > Computer Science Professor > > (on leave at INRIA from IRIF/University Paris Diderot) > > > > Director > > Software Heritage https://www.softwareheritage.org > > INRIA > > Bureau C328 E-mail : roberto at dicosmo.org > > 2, Rue Simone Iff Web page : http://www.dicosmo.org > > CS 42112 Twitter : http://twitter.com/rdicosmo > > 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 > > ------------------------------------------------------------------ > > GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 > > > From hendrik at topoi.pooq.com Thu Jan 9 09:09:45 2020 From: hendrik at topoi.pooq.com (Hendrik Boom) Date: Thu, 9 Jan 2020 09:09:45 -0500 Subject: [TYPES] [acmbulletin@acm.org: Message from the ACM President Regarding Open Access] Message-ID: <20200109140945.rjhhcmu257vtftci@topoi.pooq.com> I thought I should forward ACM's president's recent statement about this controversy, since a few days have elapsed and no one here has mentioned it. -- hendrik ----- Forwarded message from ACM Bulletins ----- Date: Tue, 7 Jan 2020 17:48:04 -0500 (EST) From: ACM Bulletins To: "Hendrik J. Boom" Subject: Message from the ACM President Regarding Open Access Message from the ACM President Regarding Open Access January 7, 2020 -------------------------------------------------------------------------------------------------------------------- Message from the ACM President Regarding Open Access Over the past few weeks, ACM leadership has listened to the concerns of our members regarding a letter we signed on to that addressed a forthcoming US Presidential Executive Order regarding the embargo of US federally-funded research. Our members have raised many important issues about the content of that letter. In response, ACM is sending a follow up letter to Office of Science and Technology Policy (OSTP) to clarify ACM's position on Open Access and its support for a sustainable approach to Open Access. It will make the points we outline below. The letter was interpreted by some ACM members as indicating that ACM is against Open Access. This could not be further from our intention. ACM chose to take the first steps in support of Open Science ideals almost a decade ago -- long before the existence of Plan S in Europe or the 2013 US OSTP Open Access Mandate. For years, ACM authors have had the right to post accepted versions of their works to non-commercial repositories (including arXiv and institutional repositories). ACM-sponsored conferences can choose to make their proceedings publicly available from their own websites, either for a limited time or permanently. ACM Special Interest Groups can choose to make the publications from all their conferences publicly available. ACM is committed to a sustainable future where all peer-reviewed scholarly articles will be Open Access. The transition to this model will take time and needs to be done in a way that ensures sustainability. Full Open Access will benefit the field of computer science significantly by increasing the sharing and citation of research accomplishments. Some of you commented on the US-centric focus of the White House directive and ACM's response. The Executive Order would only impact research supported by US federal funding. However, as a global organization ACM is also engaged with related efforts in Europe, Japan, China, and elsewhere. We regret that co-signing the letter regarding the Executive Order created confusion and concern. Our publications policies and our focus on developing sustainable publication models for Open Access are both long-standing and forward-looking. Financially "sustainable" publications models are key to ACM's future and its ability to reinvest in activities that promote the scientific foundations of computing. It is worth saying that ACM, too, had concerns about some language and the general tone of the letter, but ultimately decided that those concerns were outweighed by the risks associated with the White House issuing an Executive Order without proper consultation with stakeholders or consideration of the ramifications. In retrospect, we misgauged how our participation would be interpreted by the community. For this we are indeed sorry. We will be scheduling webinars to give members of the community an opportunity to ask questions and share opinions, and to provide information about ACM's Open Access policies, initiatives, and future directions. The schedule for these will be posted on https://www.acm.org over the coming days. Again, we greatly appreciate the comments and perspectives of our members and will continue to incorporate their feedback into our thinking and decision processes going forward. Cherri Pancake ACM President -------------------------------------------------------------------------------------------------------------------- ACM Bulletin Archive: (https://www.acm.org/membership/acm-bulletin-archive) Copyright (c) 2020, ACM, Inc. All rights reserved -------------------------------------------------------------------------------------------------------------------- You are subscribed with hendrik at TOPOI.POOQ.COM. To unsubscribe: https://optout.acm.org/unsubscribe.cfm?rm=pSfNmcLA3337893C6DC6D01ABA9E061D2F60ED379487A6E59545065554B4B7065554453&ln=BULLETIN ----- End forwarded message ----- From fritz at henglein.com Thu Jan 9 14:25:59 2020 From: fritz at henglein.com (Fritz Henglein) Date: Thu, 9 Jan 2020 20:25:59 +0100 Subject: [TYPES] ACM recants Message-ID: Hats off. https://www.acm.org/binaries/content/assets/about/acm-letter-to-ostp.pdf -- Fritz Henglein fritz at henglein.com +45-30589576 From roberto at dicosmo.org Sun Jan 12 16:16:48 2020 From: roberto at dicosmo.org (Roberto Di Cosmo) Date: Sun, 12 Jan 2020 22:16:48 +0100 Subject: [TYPES] Towards Sustainable Open Access: thoughts on the ACM OPEN transformative plan In-Reply-To: References: <20200112204727.GB10628@traveler> Message-ID: Dear all, the strong reaction to ACM signing the infamous letter from the 135 institutions confirms that in our research area we are today largely in favour of Open Access: it is not surprising considering the tradition and values of our community. The good news is that after a quarter of a century of declarations, discussions, and little progress, powerful forces are now setting tight deadlines in order to finally trigger a real transition on a global scale. In Europe, Plan S has been a strong political move, pushing a coalition of funding agencies to force 100% open acces by 2021 on publications issued by research they fund; we can expect the US proposal that sparked the infamous letter will be an equivalent strong push forward in the US. Moving from a generic support of Open Access to a rational approach to *Sustainable Open Access*, though, is more complex than it seems. - Should we go for "green open access ", i.e. self archiving the author version of our papers somewhere like we do in France with the HAL platform (that still has a cost to cover)? - Should we go for "gold open access ", aka "author pays", maybe with some discount as per SIGPLAN sponsorship? - And what about the "diamond" or "platinum" open access, where neither readers nor authors pay (rest assured, somebody *does* pay, there is no free lunch :-))? In any case, the big question is how costs should be covered, and here the debate seems mostly focused on the "right" price for publishing a single article (or APC, for article processing charge). The original version of the Plan S was strongly oriented towards gold open access ("author pays") with capped APC covered by institutions, not individuals, even if it was later clarified that green open access is also acceptable (this is called the "repository route" in II.2 of the implementation guidelines of Plan S ). Let me say upfront that I *strongly dislike* the APC approach, for a very simple reason that can be resumed in a statement that was attributed to a famous billionaire: "*If you want to get rich, build something that has a fixed cost and engenders variable income, and then get as many customers as possible*". There are indeed two main approaches to charging for an infrastructure (like a telephone network, a highway, the Internet or ... a publishing system): - the first is to charge "per use", e.g. phone calls by the minute, data per megabyte, etc., and this is how many big fortunes were made: these infrastructures have usually a fixed cost that is independent on its use, so when you have many users, the "variable income" quickly outweighs the fixed cost, and you can buy a Ferrari, a private Jet, a skyscraper, etc. - the second is to calculate the cost, add some reasonable margin for investments, and divide the result among the users (aka "mutualising costs"): this way, the more users come, the less the amount they need to pay. No Ferrari, here :-) Framing the debate in terms of the value of an APC, even capped, falls squarely in the first approach, and IMHO is a Trojan horse for large publishing corporations to keep their double digit profit margins, or even increase them, in the transition to Open Access. And those double digit profits are money that is stripped away from our global research effort! The ACM OPEN plan (https://libraries.acm.org/subscriptions-access/acmopen), on the other hand, falls squarely in the second approach, and is potentially a viable and virtuous one. I say *potentially* because, as many pointed out (and as stated in the text of the ongoing petition ), the calculations of the "cost" that is proposed to mutualise seem to include more than the publication process alone. But also because we should think at a *more global scale* and see what parts of the ACM publishing infrastructure is specific, and what part should be mutualised with other entities, bringing the overall cost down. More clarification is needed, but the recent second letter from ACM leadership lets us hope that ACM is able to listen to its members. In any case, it's important in this debate to have a clear sustainability plan, and analyze all the costs involved. On the one hand, one should not add to the bill costs unrelated to the publishing infrastructure. On the other hand, one must refrain from thinking that there is no cost apart from our own work as researchers/reviewers/editors/pc-chairs: even simply maintaining an online archive for the long term has a real, uncompressible cost, that we usually do not see until we have to actually run one [disclosure: I'm running one now :-)]. All the best -- Roberto ------------------------------------------------------------------ Computer Science Professor (on leave at INRIA from IRIF/University Paris Diderot) Director Software Heritage https://www.softwareheritage.org INRIA Bureau C328 E-mail : roberto at dicosmo.org 2, Rue Simone Iff Web page : http://www.dicosmo.org CS 42112 Twitter : http://twitter.com/rdicosmo 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 ------------------------------------------------------------------ GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 From mwh at cs.umd.edu Tue Jan 14 09:17:05 2020 From: mwh at cs.umd.edu (Michael Hicks) Date: Tue, 14 Jan 2020 09:17:05 -0500 Subject: [TYPES] Towards Sustainable Open Access: thoughts on the ACM OPEN transformative plan In-Reply-To: References: <20200112204727.GB10628@traveler> Message-ID: Hi all. This email, with a small bit of editing/link-adding, has now been turned into a blog post at PL Perspectives, the SIGPLAN blog. Please add your thoughts there! https://blog.sigplan.org/2020/01/14/what-is-a-sustainable-path-to-open-access/ Thanks, Roberto, for your thoughts about this issue! -Mike On Sun, Jan 12, 2020 at 4:49 PM Roberto Di Cosmo wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear all, > the strong reaction to ACM signing the infamous letter from the 135 > institutions > < > https://presspage-production-content.s3.amazonaws.com/uploads/1508/coalitionletteropposinglowerembargoes-864869.pdf > > > confirms > that in our research area we are today largely in favour of Open Access: it > is not surprising considering the tradition and values of our community. > The good news is that after a quarter of a century of declarations, > discussions, and little progress, powerful forces are now setting tight > deadlines in order to finally trigger a real transition on a global scale. > > In Europe, Plan S has been a strong > political move, pushing a coalition of funding agencies to force 100% open > acces by 2021 on publications issued by research they fund; we can expect > the US proposal that sparked the infamous letter will be an equivalent > strong push forward in the US. > > Moving from a generic support of Open Access to a rational approach to > *Sustainable > Open Access*, though, is more complex than it seems. > > - Should we go for "green open access > ", i.e. self > archiving the author version of our papers somewhere like we do in > France > with the HAL platform (that still has a cost to cover)? > - Should we go for "gold open access > ", aka "author > pays", > maybe with some discount as per SIGPLAN sponsorship? > - And what about the "diamond" or "platinum" open access, where neither > readers nor authors pay (rest assured, somebody *does* pay, there is no > free lunch :-))? > > In any case, the big question is how costs should be covered, and here the > debate seems mostly focused on the "right" price for publishing a single > article (or APC, for article processing charge). > > The original version of the Plan S was strongly oriented towards gold open > access ("author pays") with capped APC covered by institutions, not > individuals, even if it was later clarified that green open access is also > acceptable (this is called the "repository route" in II.2 of the > implementation guidelines of Plan S > < > https://www.coalition-s.org/addendum-to-the-coalition-s-guidance-on-the-implementation-of-plan-s/principles-and-implementation/ > > > ). > > Let me say upfront that I *strongly dislike* the APC approach, for a very > simple reason that can be resumed in a statement that was attributed to a > famous billionaire: "*If you want to get rich, build something that has a > fixed cost and engenders variable income, and then get as many customers as > possible*". > > There are indeed two main approaches to charging for an infrastructure > (like a telephone network, a highway, the Internet or ... a publishing > system): > > - the first is to charge "per use", e.g. phone calls by the minute, data > per megabyte, etc., and this is how many big fortunes were made: these > infrastructures have usually a fixed cost that is independent on its > use, > so when you have many users, the "variable income" quickly outweighs the > fixed cost, and you can buy a Ferrari, a private Jet, a skyscraper, etc. > - the second is to calculate the cost, add some reasonable margin for > investments, and divide the result among the users (aka "mutualising > costs"): this way, the more users come, the less the amount they need to > pay. No Ferrari, here :-) > > Framing the debate in terms of the value of an APC, even capped, falls > squarely in the first approach, and IMHO is a Trojan horse for large > publishing corporations to keep their double digit profit margins, or even > increase them, in the transition to Open Access. > And those double digit profits are money that is stripped away from our > global research effort! > > The ACM OPEN plan (https://libraries.acm.org/subscriptions-access/acmopen > ), > on the other hand, falls squarely in the second approach, and is > potentially a viable and virtuous one. I say *potentially* because, as many > pointed out (and as stated in the text of the ongoing petition > < > https://www.change.org/p/association-for-computing-machinery-acm-support-open-access > >), > the calculations of the "cost" that is proposed to mutualise seem to > include more than the publication process alone. > But also because we should think at a *more global scale* and see what > parts of the ACM publishing infrastructure is specific, and what part > should be mutualised with other entities, bringing the overall cost down. > More clarification is needed, but the recent second letter from ACM > leadership > > lets us hope that ACM is able to listen to its members. > > In any case, it's important in this debate to have a clear sustainability > plan, and analyze all the costs involved. On the one hand, one should not > add to the bill costs unrelated to the publishing infrastructure. On the > other hand, one must refrain from thinking that there is no cost apart from > our own work as researchers/reviewers/editors/pc-chairs: even simply > maintaining an online archive for the long term has a real, uncompressible > cost, that we usually do not see until we have to actually run one > [disclosure: I'm running one now :-)]. > > All the best > > -- > Roberto > > ------------------------------------------------------------------ > Computer Science Professor > (on leave at INRIA from IRIF/University Paris Diderot) > > Director > Software Heritage https://www.softwareheritage.org > INRIA > Bureau C328 E-mail : roberto at dicosmo.org > 2, Rue Simone Iff Web page : http://www.dicosmo.org > CS 42112 Twitter : http://twitter.com/rdicosmo > 75589 Paris Cedex 12 Tel : +33 1 80 49 44 42 > ------------------------------------------------------------------ > > GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 > From might at cs.utah.edu Thu Jan 23 14:23:18 2020 From: might at cs.utah.edu (Matt Might) Date: Thu, 23 Jan 2020 13:23:18 -0600 Subject: [TYPES] Signable open letter to OSTP regarding open access Message-ID: Following up on prior discussions around open access, if you'd like to send a message to OSTP that you support open access, we've been collecting signatures on an open letter here: http://www.oaintheusa.com/ I will make sure that the key players within OSTP see this. I'm hoping for 10,000 signatures. This has been in the works for a while, and there is momentum building. There's a nontrivial chance we can pull this off. Thanks for considering signing, and if you can spread this, please do! Matt From julbinb at gmail.com Thu Jan 23 19:19:29 2020 From: julbinb at gmail.com (Julia Belyakova) Date: Thu, 23 Jan 2020 19:19:29 -0500 Subject: [TYPES] Signable open letter to OSTP regarding open access In-Reply-To: References: Message-ID: Thank you for sharing! Can any person working in the US sign the letter? Or is it only for Amrican citizens/residents? -- Kind regards, Julia ??, 23 ???. 2020 ?. ? 16:55, Matt Might : > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Following up on prior discussions around open access, if you'd like to > send a message to OSTP that you support open access, we've been > collecting signatures on an open letter here: > > http://www.oaintheusa.com/ > > I will make sure that the key players within OSTP see this. I'm hoping > for 10,000 signatures. > > This has been in the works for a while, and there is momentum > building. There's a nontrivial chance we can pull this off. > > Thanks for considering signing, and if you can spread this, please do! > > Matt > From might at cs.utah.edu Thu Jan 23 23:14:55 2020 From: might at cs.utah.edu (Matt Might) Date: Thu, 23 Jan 2020 22:14:55 -0600 Subject: [TYPES] Signable open letter to OSTP regarding open access In-Reply-To: References: Message-ID: Any person living or working in the US can sign, regardless of citizenship. US citizens abroad may also sign. Thanks! On Thu, Jan 23, 2020 at 6:19 PM Julia Belyakova wrote: > Thank you for sharing! > > Can any person working in the US sign the letter? Or is it only for > Amrican citizens/residents? > > -- > Kind regards, Julia > > > ??, 23 ???. 2020 ?. ? 16:55, Matt Might : > >> [ The Types Forum, >> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > >> >> Following up on prior discussions around open access, if you'd like to >> send a message to OSTP that you support open access, we've been >> collecting signatures on an open letter here: >> >> http://www.oaintheusa.com/ >> >> I will make sure that the key players within OSTP see this. I'm hoping >> for 10,000 signatures. >> >> This has been in the works for a while, and there is momentum >> building. There's a nontrivial chance we can pull this off. >> >> Thanks for considering signing, and if you can spread this, please do! >> >> Matt >> > -- Sent from Gmail Mobile From kareem.ergawy at gmail.com Wed Feb 5 10:26:36 2020 From: kareem.ergawy at gmail.com (Kareem Ergawy) Date: Wed, 5 Feb 2020 16:26:36 +0100 Subject: [TYPES] [newbie contribution] C++ implementations + tests for TAPL Message-ID: Hello, Sort of a newbie contribution, hopefully this is interesting for others in the forum. I am currently self-studying "Types and Programming Languages" book and decided to provide mature implementations of the languages and type systems studied in the book in C++. Here is a link to the repo: https://github.com/KareemErgawy/types-and-programming-languages. Each implementation contains a large battery of tests and hopefully the code makes it easy to add new tests. I am already a full-time software engineer (general C++ development) hoping to find a compilers-related job in the future and TAPL got me excited to learn more about type systems. Hopefully, this is not straying too much from the purpose of the forums and that folks here are interested in such newbie contributions :). Best, Kareem From tringer at cs.washington.edu Mon Feb 17 07:39:47 2020 From: tringer at cs.washington.edu (Talia Ringer) Date: Mon, 17 Feb 2020 04:39:47 -0800 Subject: [TYPES] Question about UIP and decidable equality Message-ID: Hello, I have a question that has been bugging me for a long time, and I'm hoping someone has some insight. I'm working in vanilla CIC, and right now I'm interested in types "A : Type" and "B : I -> Type" for which there is a function "indexer" that makes the following type equivalence hold: "forall i : I, equiv { a : A | indexer a = i} (B i)" Everyone's favorite example of this is a list, a vector, and the length function. I know for a fact that when the type "I" of the index "i" has decidable equality, without any axioms, I can derive that UIP holds on that index type. That's true in the vector case, for example. Where I'm stuck is that I sometimes have types for which equality on "I" is in general not decidable, but for which I can choose some "J" with decidable equality and some "C : J -> Type" such that: "equiv (sigma (i : I) . B i) (sigma (j : J) . C j)" This means it's very easy to work over C, since I get UIP over the index. But I'm not interested in functions and theorems that are defined over C. I would like them over B. And I would like for equalities of equalities of the indices of B to not be painful. Is there anything that I can say, without any additional axioms, about how equalities between the indices of B relate to one another? That is, can I get a result similar to UIP holding on the indices of B, even if it does not hold in general on I? If so, how? If not, why not, and what additional axioms would be necessary? Thanks! I'm happy to clarify if anyone is confused about any details; it was hard to make this concise. Talia From Isaac.Oscar.Gariano at ecs.vuw.ac.nz Mon Feb 17 20:13:05 2020 From: Isaac.Oscar.Gariano at ecs.vuw.ac.nz (Isaac) Date: Tue, 18 Feb 2020 01:13:05 +0000 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: Hello, I am currently working on writing a type-checker for an object-oriented structural type-system with generics and recursive types. In particular, it's supports recursive parametric type alias, e.g: type List[X] = { head -> X; tail -> List[X]} Then, if T is a type, a "List[T]" is any object with a head method that returns "T"s and a tail method that returns "List[T]"s. Note that partially-applied type aliases are not allowed, e.g. "List" is not a valid type. Type-alias can have any number of type-parameters, including zero, and can also be mutually recursive. Now since this type-system is structurally typed, if I define: type Integer-List = { head -> Integer; tail -> Integer-List; sum -> Integer } I should have that Integer-List is a sub-type of List[Integer] (which I will denote Integer-List <: List[Integer]) In this simple case, I can use a standard con-inductive algorithm for recursive types: - Integer-List <: List[Integer]? - { head -> Integer; tail -> Integer-List; sum -> Integer } <: {head -> Integer; tail -> List[Integer]}? - Integer <: Integer? and Integer-List <: List[Integer]? - Yes (by reflexivity) and yes (by co-induction, since we are already trying to work out if Integer-List <: List[Integer]) Ok, now on to the hard bit, what if I have: type Exploding-List[X] = { head -> X; explode -> Exploding-List[Exploding-List[X]] } type Integer-Explosion = { head-> Integer; explode -> Exploding-List[Integer-Explosion] } Is Integer-Explosion a sub-type of Exploding-List[Integer]? - Integer-Explosion <: Exploding-List[Integer]? - { head-> Integer; explode -> Exploding-List[Integer-Explosion] } <: { head -> Integer; explode -> Exploding-List[Exploding-List[Integer]] }? - Integer <: Integer? and Exploding-List[Integer-Explosion] <: Exploding-List[Exploding-List[Integer]]? - yes (by reflexivity) and { head -> Integer-Explosion; explode -> Exploding-List[Exploding-List[Integer-Explosion]] } <: { head -> Exploding-List[Integer]; explode -> Exploding-List[Exploding-List[Exploding-List[Integer]]] } - Integer-Explosion <: Exploding-List[Integer] and Exploding-List[Exploding-List[Integer-Explosion]] <: Exploding-List[Exploding-List[Exploding-List[Integer]]] - yes (by co-induction) and .... If I keep going I will get increasingly large sub-type tests of form Exploding-List[... Exploding-List[Integer-Explosion] ... ] <: Exploding-List[...Exploding-List[Integer]...]. Co-induction wont save me here as there are an infinite number of such sub-type tests I need to perform (since there are an infinite number of types). The question is, is anyone aware of any algorithm I could use to compute such sub-typing relations? or does anyone know of an existing type-checker that already does this? After some research, the only language I could find that appears to support such constructs is the one from this paper (which they call ?-rec-abs) https://link.springer.com/content/pdf/10.1007%2F978-3-642-39212-2_28.pdf. However they do not discuss sub-typing. I find this particularly interesting as I don't think that such infinitely expanding sub-type tests are that pathological, for example it might be reasonable to have a List[T] type with a split(T) -> List[List[T]] method: the idea being that list.split(x) will split the list on occurrences of "x" (like you can do with strings in many languages: "1,2,3".split(",") = ["1", "2", "3"]. I'm not aware of any real-world programming languages that support structural typing like the above, I've tried Scala but it is not possible to have a recursive type alias, and OCaml, which only allows type alias with simple recursion like my original "List" example, and flat out rejects my Exploding-List type. ? Isaac Oscar Gariano? From moez at cs.rice.edu Tue Feb 18 01:13:12 2020 From: moez at cs.rice.edu (moez at cs.rice.edu) Date: Tue, 18 Feb 2020 06:13:12 +0000 (UTC) Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? Message-ID: <5E7984F788544BED.c537341d-4de6-40ac-b5cf-c3234620c2d2@mail.outlook.com> Hello, Have you checked Kennedy and Pierce's paper "On Decidability of Nominal Subtyping with Variance"? -Moez From: Types-list on behalf of Isaac Sent: Tuesday, February 18, 2020, 07:49 To: types-list at lists.seas.upenn.edu Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] Hello, I am currently working on writing a type-checker for an object-oriented structural type-system with generics and recursive types. In particular, it's supports recursive parametric type alias, e.g: type List[X] = { head -> X; tail -> List[X]} Then, if T is a type, a "List[T]" is any object with a head method that returns "T"s and a tail method that returns "List[T]"s. Note that partially-applied type aliases are not allowed, e.g. "List" is not a valid type. Type-alias can have any number of type-parameters, including zero, and can also be mutually recursive. Now since this type-system is structurally typed, if I define: type Integer-List = { head -> Integer; tail -> Integer-List; sum -> Integer } I should have that Integer-List is a sub-type of List[Integer] (which I will denote Integer-List <: List[Integer]) In this simple case, I can use a standard con-inductive algorithm for recursive types: - Integer-List <: List[Integer]? - { head -> Integer; tail -> Integer-List; sum -> Integer } <: {head -> Integer; tail -> List[Integer]}? - Integer <: Integer? and Integer-List <: List[Integer]? - Yes (by reflexivity) and yes (by co-induction, since we are already trying to work out if Integer-List <: List[Integer]) Ok, now on to the hard bit, what if I have: type Exploding-List[X] = { head -> X; explode -> Exploding-List[Exploding-List[X]] } type Integer-Explosion = { head-> Integer; explode -> Exploding-List[Integer-Explosion] } Is Integer-Explosion a sub-type of Exploding-List[Integer]? - Integer-Explosion <: Exploding-List[Integer]? - { head-> Integer; explode -> Exploding-List[Integer-Explosion] } <: { head -> Integer; explode -> Exploding-List[Exploding-List[Integer]] }? - Integer <: Integer? and Exploding-List[Integer-Explosion] <: Exploding-List[Exploding-List[Integer]]? - yes (by reflexivity) and { head -> Integer-Explosion; explode -> Exploding-List[Exploding-List[Integer-Explosion]] } <: { head -> Exploding-List[Integer]; explode -> Exploding-List[Exploding-List[Exploding-List[Integer]]] } - Integer-Explosion <: Exploding-List[Integer] and Exploding-List[Exploding-List[Integer-Explosion]] <: Exploding-List[Exploding-List[Exploding-List[Integer]]] - yes (by co-induction) and .... If I keep going I will get increasingly large sub-type tests of form Exploding-List[... Exploding-List[Integer-Explosion] ... ] <: Exploding-List[...Exploding-List[Integer]...]. Co-induction wont save me here as there are an infinite number of such sub-type tests I need to perform (since there are an infinite number of types). The question is, is anyone aware of any algorithm I could use to compute such sub-typing relations? or does anyone know of an existing type-checker that already does this? After some research, the only language I could find that appears to support such constructs is the one from this paper (which they call ?-rec-abs) https://link.springer.com/content/pdf/10.1007%2F978-3-642-39212-2_28.pdf. However they do not discuss sub-typing. I find this particularly interesting as I don't think that such infinitely expanding sub-type tests are that pathological, for example it might be reasonable to have a List[T] type with a split(T) -> List[List[T]] method: the idea being that list.split(x) will split the list on occurrences of "x" (like you can do with strings in many languages: "1,2,3".split(",") = ["1", "2", "3"]. I'm not aware of any real-world programming languages that support structural typing like the above, I've tried Scala but it is not possible to have a recursive type alias, and OCaml, which only allows type alias with simple recursion like my original "List" example, and flat out rejects my Exploding-List type. ? Isaac Oscar Gariano? From amintimany at gmail.com Tue Feb 18 04:28:06 2020 From: amintimany at gmail.com (Amin Timany) Date: Tue, 18 Feb 2020 10:28:06 +0100 Subject: [TYPES] Question about UIP and decidable equality In-Reply-To: References: Message-ID: <8E7157B7-2725-413D-8706-5F07BD90640B@gmail.com> Hi Talia, In general the situation that you are describing here between I, B, J and C does not seem to be strong enough to derive any interesting properties about equalities in I. Consider the following example: I = Type B(A : I) = sigma (n : nat) . (Vec A n) J = nat C(n : J) = sigma (A : Type) . (Vec A n) Here "equiv (sigma (i : I) . B i) (sigma (j : J) . C j)? holds trivially but that does not say anything about equalities in I. Maybe some constraint on I, B, J, and C would make it possible to say something about equalities in I in the specific case(s) you are working with. One specific case that comes to mind would be if there is a bijection between f : I -> J such that forall i : I, equiv B(i) C(f(i)). However in that case I would have decidable equality (and hence UIP) if J has decidable equality. So it is probably not very useful for you. Regards, Amin > On 17 Feb 2020, at 13:39, Talia Ringer wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Hello, > > I have a question that has been bugging me for a long time, and I'm hoping > someone has some insight. I'm working in vanilla CIC, and right now I'm > interested in types "A : Type" and "B : I -> Type" for which there is a > function "indexer" that makes the following type equivalence hold: > > "forall i : I, equiv { a : A | indexer a = i} (B i)" > > Everyone's favorite example of this is a list, a vector, and the length > function. > > I know for a fact that when the type "I" of the index "i" has decidable > equality, without any axioms, I can derive that UIP holds on that index > type. That's true in the vector case, for example. > > Where I'm stuck is that I sometimes have types for which equality on "I" is > in general not decidable, but for which I can choose some "J" with > decidable equality and some "C : J -> Type" such that: > > "equiv (sigma (i : I) . B i) (sigma (j : J) . C j)" > > This means it's very easy to work over C, since I get UIP over the index. > But I'm not interested in functions and theorems that are defined over C. I > would like them over B. And I would like for equalities of equalities of > the indices of B to not be painful. > > Is there anything that I can say, without any additional axioms, about how > equalities between the indices of B relate to one another? That is, can I > get a result similar to UIP holding on the indices of B, even if it does > not hold in general on I? If so, how? If not, why not, and what additional > axioms would be necessary? > > Thanks! I'm happy to clarify if anyone is confused about any details; it > was hard to make this concise. > > Talia From lepigre at mpi-sws.org Tue Feb 18 05:14:59 2020 From: lepigre at mpi-sws.org (Rodolphe Lepigre) Date: Tue, 18 Feb 2020 11:14:59 +0100 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: <20200218101459.GA42143@HPArchRod.localdomain> Hi, Together with Christophe Raffalli, we defined a language called SubML that can do more or less what you ask for. Here are a couple of links: - https://rlepigre.github.io/subml/ (online interpreter to play around), - https://github.com/rlepigre/subml/ (repository), - https://dl.acm.org/doi/10.1145/3285955 (TOPLAS paper, also on my web page). When I say "more or less", I mean that we cannot directly express your example in full. However, I see no reason why our technique could not apply given some adaptation work. What is missing from SubML to handle your example is that the type parameters of an inductive or coinductive type is fixed in its definition due to the way they are defined (using least and greatest fixpoints). The best I can do currently is: ========== include "nat.typ" type List(X) = ?List .{ head : X ; tail : List ; ? } type NatList = ?NatList.{ head : Nat ; tail : NatList ; ? } check NatList ? List(Nat) check List(Nat) ? NatList check { head : Nat ; tail : NatList ; sum : Nat ; ? } ? { head : Nat ; tail : List(Nat) ; ? } check Nat ? Nat (* Note the "List" here. *) type ExplList(X) = ?ExplList.{ head : X ; explode : List(ExplList) ; ? } type NatExpl = ?NatExpl .{ head : Nat ; explode : List(NatExpl ) ; ? } check NatExpl ? ExplList(Nat) check { head : Nat ; explode : ExplList(NatExpl) ; ? } ? { head : Nat ; explode : ExplList(ExplList(Nat)) ; ? } check ExplList(NatExpl) ? ExplList(ExplList(Nat)) check { head : Nat ; explode : ExplList(ExplList(NatExpl)) ; ? } ? { head : Nat ; explode : ExplList(ExplList(ExplList(Nat))) ; ? } check ExplList(ExplList(NatExpl)) ? ExplList(ExplList(ExplList(Nat))) ========== In fact, we can express your example in PML (https://github.com/rlepigre/pml), which is a proof system built using similar techniques (with extensions). But currently these features are quite experimental so the subtyping relations you want for your exploding lists cannot be proved by the system (including those that should be trivial...). I'll investigate when I have a moment. ========== include lib.nat type corec list?a:?? = { head : a ; tail : list?a? ; ? } type corec natlist = { head : nat ; tail : natlist ; ? } assert natlist ? list?nat? assert list?nat? ? natlist assert { head : nat ; tail : natlist ; sum : nat ; ? } ? { head : nat ; tail : list?nat? ; ? } assert nat ? nat type corec any_expl?a:?? = { head : a ; explode : any_expl?any_expl?a ?? ; ? } type corec nat_expl = { head : nat ; explode : any_expl?any_expl?nat?? ; ? } // FIXME do not work //assert any_expl?nat? ? nat_expl // should be trivial //assert nat_expl ? any_expl?nat? // should be trivial //assert { head : nat ; explode : any_expl?nat_expl? ; ? } ? { head : nat ; explode : any_expl?any_expl?nat?? } //assert any_expl?nat_expl? ? any_expl?any_expl?nat?? //assert { head : nat ; explode : any_expl?any_expl?nat_expl??; ? } ? { head : nat ; explode : any_expl?any_expl?any_expl?nat??? } //assert any_expl?nat_expl? ? any_expl?any_expl?nat?? ========== Now in terms of technique, what we do is (roughly): 1) build cyclic subtyping proofs, and 2) prove that the cyclic proof are well-founded. To do that, our (inductive and coinductive) types are annotated with ordinals, and we then check that there is some decrease in the ordinals along the cyclic structure of the proof (in our work this is done using a variation of the size change principle of Lee, Jones and Ben-Amram). Annotating types using sizes is usually called "sized types" (they are often natural numbers in the literature but sometimes also ordinals). Of course, size annotations are not always given by the user explicitly. For subtyping I think they can always be automatically inserted. However the size algebra that we currently use may not be expressive enough (it is very minima and only has variables, successor and a large enough constant to make all fixed points converge). Extensions are possible. The SubML language relies on this cyclic proof mechanism for subtyping and for termination checking, but it is certainly possible to only do the former. I hope this helps, do not hesitate to ask me for more details. Cheers, Rodolphe. -- Rodolphe Lepigre Max Planck Institute for Software Systems, Saarbr?cken, Germany https://lepigre.fr On 18/02/20 01:13, Isaac wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Hello, > > I am currently working on writing a type-checker for an object-oriented structural type-system with generics and recursive types. In particular, it's supports recursive parametric type alias, e.g: > type List[X] = { head -> X; tail -> List[X]} > > Then, if T is a type, a "List[T]" is any object with a head method that returns "T"s and a tail method that returns "List[T]"s. > Note that partially-applied type aliases are not allowed, e.g. "List" is not a valid type. > Type-alias can have any number of type-parameters, including zero, and can also be mutually recursive. > > Now since this type-system is structurally typed, if I define: > type Integer-List = { head -> Integer; tail -> Integer-List; sum -> Integer } > > I should have that Integer-List is a sub-type of List[Integer] (which I will denote Integer-List <: List[Integer]) > In this simple case, I can use a standard con-inductive algorithm for recursive types: > - Integer-List <: List[Integer]? > - { head -> Integer; tail -> Integer-List; sum -> Integer } <: {head -> Integer; tail -> List[Integer]}? > - Integer <: Integer? and Integer-List <: List[Integer]? > - Yes (by reflexivity) and yes (by co-induction, since we are already trying to work out if Integer-List <: List[Integer]) > > Ok, now on to the hard bit, what if I have: > type Exploding-List[X] = { head -> X; explode -> Exploding-List[Exploding-List[X]] } > type Integer-Explosion = { head-> Integer; explode -> Exploding-List[Integer-Explosion] } > > Is Integer-Explosion a sub-type of Exploding-List[Integer]? > - Integer-Explosion <: Exploding-List[Integer]? > - { head-> Integer; explode -> Exploding-List[Integer-Explosion] } <: { head -> Integer; explode -> Exploding-List[Exploding-List[Integer]] }? > - Integer <: Integer? and Exploding-List[Integer-Explosion] <: Exploding-List[Exploding-List[Integer]]? > - yes (by reflexivity) and > { head -> Integer-Explosion; explode -> Exploding-List[Exploding-List[Integer-Explosion]] } > <: { head -> Exploding-List[Integer]; explode -> Exploding-List[Exploding-List[Exploding-List[Integer]]] } > - Integer-Explosion <: Exploding-List[Integer] and Exploding-List[Exploding-List[Integer-Explosion]] <: Exploding-List[Exploding-List[Exploding-List[Integer]]] > - yes (by co-induction) and .... > > If I keep going I will get increasingly large sub-type tests of form Exploding-List[... Exploding-List[Integer-Explosion] ... ] <: Exploding-List[...Exploding-List[Integer]...]. > Co-induction wont save me here as there are an infinite number of such sub-type tests I need to perform (since there are an infinite number of types). > > The question is, is anyone aware of any algorithm I could use to compute such sub-typing relations? or does anyone know of an existing type-checker that already does this? > > After some research, the only language I could find that appears to support such constructs is the one from this paper (which they call ?-rec-abs) https://link.springer.com/content/pdf/10.1007%2F978-3-642-39212-2_28.pdf. > However they do not discuss sub-typing. > > I find this particularly interesting as I don't think that such infinitely expanding sub-type tests are that pathological, for example it might be reasonable to have a List[T] type with a split(T) -> List[List[T]] method: the idea being that list.split(x) will split the list on occurrences of "x" (like you can do with strings in many languages: "1,2,3".split(",") = ["1", "2", "3"]. > > I'm not aware of any real-world programming languages that support structural typing like the above, I've tried Scala but it is not possible to have a recursive type alias, and OCaml, which only allows type alias with simple recursion like my original "List" example, and flat out rejects my Exploding-List type. > > ? Isaac Oscar Gariano From francois.pottier at inria.fr Tue Feb 18 06:14:23 2020 From: francois.pottier at inria.fr (=?UTF-8?Q?Fran=c3=a7ois_Pottier?=) Date: Tue, 18 Feb 2020 12:14:23 +0100 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: Hello, If I am not mistaken, already deciding the *equivalence* of two types in the presence of recursive and parameterized type abbreviations is extremely costly: Marvin Solomon proved it equivalent to the DPDA equivalence problem ("Type definitions with parameters", POPL'78). -- Fran?ois Pottier francois.pottier at inria.fr http://cambium.inria.fr/~fpottier/ From p.giarrusso at gmail.com Tue Feb 18 09:14:52 2020 From: p.giarrusso at gmail.com (Paolo Giarrusso) Date: Tue, 18 Feb 2020 15:14:52 +0100 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: Hello, On Tue, 18 Feb 2020 at 14:36, Fran?ois Pottier wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > If I am not mistaken, already deciding the *equivalence* of two types > in the presence of recursive and parameterized type abbreviations is > extremely costly: Marvin Solomon proved it equivalent to the DPDA > equivalence problem ("Type definitions with parameters", POPL'78). > Indeed; Se?nizergues later discovered this is decidable [1] but super-exponential, and this applies already to *nested datatypes* like Exploding-List ([2, Sec. 3.4.3]); I'm not familiar with the decision algorithm, but nobody describes it as practical in this context [2, 3]. I wonder if subtyping would reduce to DPDA containment, similarly to Solomon's result, and whether *that* problem is decidable. If you take F_{\omega\mu} but restrict fixpoints to result kind * (which excludes nested datatypes), definable types are equivalent to regular trees? so equivalence can be decided [2] by an extension of Amadio-Cardelli's algorithm that normalizes types first, as earlier conjectured by Fran?ois Pottier [3]. That paper only deals with type equivalence, but Amadio-Cardelli's algorithm decides equirecursive subtyping; a combination of the two algorithms should decide F_{\omega\mu<:}. That appears to be equivalent to what you Isaac are doing (tho I suspect the "standard coinductive subtyping" is the Brandt-Henglein reformulation [4]), but I do not think anybody worked out the details and especially correctness/type soundness proofs? [1] G?raud S?nizergues. ?Some Applications of the Decidability of DPDA?s Equivalence.? https://doi.org/10.1007/3-540-45132-3_7. Corollary 6. [2] Yufei Cai, Paolo G. Giarrusso, and Klaus Ostermann. ?System F-Omega with Equirecursive Types for Datatype-Generic Programming.? http://dl.acm.org/citation.cfm?id=2837660. [3] Fran?ois Pottier. http://lists.seas.upenn.edu/pipermail/types-list/2011/001525.html [4] Michael Brandt and Fritz Henglein. ?Coinductive Axiomatization of Recursive Type Equality and Subtyping.? http://dl.acm.org/citation.cfm?id=2379036.2379037. Cheers, -- Paolo G. Giarrusso From abadi at cs.ucsc.edu Tue Feb 18 09:15:32 2020 From: abadi at cs.ucsc.edu (Martin Abadi) Date: Tue, 18 Feb 2020 11:15:32 -0300 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: Hello, TypeScript has parameterized recursive definitions and structural subtyping. The language has evolved over time, and I am afraid I have not looked at the current definitions, but, early on, the designers were aware of some of the algorithmic difficulties that Fran?ois Pottier mentions and introduced restrictions to avoid them. I don't know whether the examples of interest to Isaac Oscar Gariano would be expressible. Regards, Martin On Tue, Feb 18, 2020 at 10:36 AM Fran?ois Pottier wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > > Hello, > > If I am not mistaken, already deciding the *equivalence* of two types > in the presence of recursive and parameterized type abbreviations is > extremely costly: Marvin Solomon proved it equivalent to the DPDA > equivalence problem ("Type definitions with parameters", POPL'78). > > -- > Fran?ois Pottier > francois.pottier at inria.fr > http://cambium.inria.fr/~fpottier/ > From blichtman623 at gmail.com Tue Feb 18 17:14:38 2020 From: blichtman623 at gmail.com (Benjamin Lichtman) Date: Tue, 18 Feb 2020 14:14:38 -0800 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: Message-ID: Hello all, TypeScript team member here. This example actually is supported by TypeScript. See here for a demonstration. Best, Ben On Tue, Feb 18, 2020 at 1:14 PM Martin Abadi wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Hello, > > TypeScript has parameterized recursive > definitions and structural subtyping. The language has evolved over time, > and I am afraid I have not looked at the current definitions, but, early > on, the designers were aware of some of the algorithmic difficulties > that Fran?ois Pottier mentions and introduced restrictions to avoid them. I > don't know whether the examples of interest to Isaac Oscar Gariano would be > expressible. > > Regards, > Martin > > > On Tue, Feb 18, 2020 at 10:36 AM Fran?ois Pottier < > francois.pottier at inria.fr> > wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > > > Hello, > > > > If I am not mistaken, already deciding the *equivalence* of two types > > in the presence of recursive and parameterized type abbreviations is > > extremely costly: Marvin Solomon proved it equivalent to the DPDA > > equivalence problem ("Type definitions with parameters", POPL'78). > > > > -- > > Fran?ois Pottier > > francois.pottier at inria.fr > > http://cambium.inria.fr/~fpottier/ > > > From might at cs.utah.edu Sun Feb 23 21:43:46 2020 From: might at cs.utah.edu (Matt Might) Date: Sun, 23 Feb 2020 20:43:46 -0600 Subject: [TYPES] Signable open letter to OSTP regarding open access In-Reply-To: References: Message-ID: As a further follow-up, the petition succeeded, and now OSTP has opened a formal comments period regarding a mandate for open access to all federally funded research. Currently, the lobbyists for the publishers are flooding OSTP political appointees with endless objections. They are also claiming there is no support for this from rank and file scientists. If you have the time, sending in a response from you and/or your institution would be very valuable at this critical moment in the internal policy debate: https://www.federalregister.gov/documents/2020/02/19/2020-03189/request-for-information-public-access-to-peer-reviewed-scholarly-publications-data-and-code Please note that OSTP is also considering requiring that code and data necessary to reproduce results be made available as well, so feel free to weigh in on that as well. Thank you, Matt On Thu, Jan 23, 2020 at 1:23 PM Matt Might wrote: > > Following up on prior discussions around open access, if you'd like to > send a message to OSTP that you support open access, we've been > collecting signatures on an open letter here: > > http://www.oaintheusa.com/ > > I will make sure that the key players within OSTP see this. I'm hoping > for 10,000 signatures. > > This has been in the works for a while, and there is momentum > building. There's a nontrivial chance we can pull this off. > > Thanks for considering signing, and if you can spread this, please do! > > Matt From Isaac.Oscar.Gariano at ecs.vuw.ac.nz Mon Feb 24 03:03:08 2020 From: Isaac.Oscar.Gariano at ecs.vuw.ac.nz (Isaac) Date: Mon, 24 Feb 2020 08:03:08 +0000 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: , Message-ID: Thanks Paolo and Fran?ois, From my understanding of Solomon's paper, it seems that the problem of sub-typing would be equivalent to containment of deterministic context free languages, unfortunately this paper https://www.sciencedirect.com/science/article/pii/S0019995866800190 has proven it undecidable. That being said, I would be happy with a sound algorithm that works for reasonable practical cases (like my exploding list). ? Isaac Oscar Gariano? ________________________________ From: Paolo Giarrusso Sent: 19 February 2020 3:14 AM To: Isaac ; Types list Subject: Re: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? Hello, On Tue, 18 Feb 2020 at 14:36, Fran?ois Pottier > wrote: [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] If I am not mistaken, already deciding the *equivalence* of two types in the presence of recursive and parameterized type abbreviations is extremely costly: Marvin Solomon proved it equivalent to the DPDA equivalence problem ("Type definitions with parameters", POPL'78). Indeed; Se?nizergues later discovered this is decidable [1] but super-exponential, and this applies already to *nested datatypes* like Exploding-List ([2, Sec. 3.4.3]); I'm not familiar with the decision algorithm, but nobody describes it as practical in this context [2, 3]. I wonder if subtyping would reduce to DPDA containment, similarly to Solomon's result, and whether *that* problem is decidable. If you take F_{\omega\mu} but restrict fixpoints to result kind * (which excludes nested datatypes), definable types are equivalent to regular trees? so equivalence can be decided [2] by an extension of Amadio-Cardelli's algorithm that normalizes types first, as earlier conjectured by Fran?ois Pottier [3]. That paper only deals with type equivalence, but Amadio-Cardelli's algorithm decides equirecursive subtyping; a combination of the two algorithms should decide F_{\omega\mu<:}. That appears to be equivalent to what you Isaac are doing (tho I suspect the "standard coinductive subtyping" is the Brandt-Henglein reformulation [4]), but I do not think anybody worked out the details and especially correctness/type soundness proofs? [1] G?raud S?nizergues. ?Some Applications of the Decidability of DPDA?s Equivalence.? https://doi.org/10.1007/3-540-45132-3_7. Corollary 6. [2] Yufei Cai, Paolo G. Giarrusso, and Klaus Ostermann. ?System F-Omega with Equirecursive Types for Datatype-Generic Programming.? http://dl.acm.org/citation.cfm?id=2837660. [3] Fran?ois Pottier. http://lists.seas.upenn.edu/pipermail/types-list/2011/001525.html [4] Michael Brandt and Fritz Henglein. ?Coinductive Axiomatization of Recursive Type Equality and Subtyping.? http://dl.acm.org/citation.cfm?id=2379036.2379037. Cheers, -- Paolo G. Giarrusso From Isaac.Oscar.Gariano at ecs.vuw.ac.nz Mon Feb 24 03:16:21 2020 From: Isaac.Oscar.Gariano at ecs.vuw.ac.nz (Isaac) Date: Mon, 24 Feb 2020 08:16:21 +0000 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: , Message-ID: Dear Stefan, Thanks for the reply! >> Note that partially-applied type aliases are not allowed, e.g. "List" is not a valid type. > I'm not completely sure what you mean by "valid type" here. > Clearly, `List` is not a proper type, so do you mean that `list` cannot > appear un-applied within a type expression? Can arguments to your type > aliases be something else than proper types? No they cannot. Any mention of a type-alias name must be supplied with the appropriate number of type arguments or it is an error. (alternatively, the missing ones might be inferred). In particular, higher-order type aliases are not allowed: type Foo[C] = C[Int] // error, only type-aliases can be supplied with paramaters type Bar[X] = { head -> X } type Baz = Foo[Bar] // error, Bar needs 1 type argument Thus all type-aliases are of kind * -> * -> ... -> * (for some finite number of arrows). Although I believe languages like Haskell or Scala can support such constructs. > AFAICT here you need to use the property: > ?X,Y. X <: Y implies Exploding-List[X] <: Exploding-List[Y] > in order to prove the subtyping without hitting the inf-loop. > If all your arguments are proper types, than you can probably get >a working algorithm that keeps track of the polarity of arguments and > uses this kind of reasoning (instead of expanding the definition) > whenever faced with > Exploding-List[...] <: Exploding-List[...] > Now I don't think it's going to help you very much with This looks like inferring the variance of the type-paramaters? (which Typescript appears to be doing). I'll look into literature on this, thanks. > Exploding1-List[...] <: Exploding2-List[...] > and other fun stuff. Sadly not, but perhaps that is an unlikely case? Although I feel rejecting such code kind of defeats the spirit of a structural type system. From spam.caminati at gmail.com Mon Feb 24 08:46:52 2020 From: spam.caminati at gmail.com (marco caminati) Date: Mon, 24 Feb 2020 13:46:52 +0000 Subject: [TYPES] Signable open letter to OSTP regarding open access In-Reply-To: References: Message-ID: On 2/24/20, Matt Might wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > As a further follow-up, the petition succeeded, and now OSTP has > opened a formal comments period regarding a mandate for open access to > all federally funded research. > > Currently, the lobbyists for the publishers are flooding OSTP > political appointees with endless objections. > > They are also claiming there is no support for this from rank and file > scientists. > > If you have the time, sending in a response from you and/or your > institution would be very valuable at this critical moment in the > internal policy debate: > > https://www.federalregister.gov/documents/2020/02/19/2020-03189/request-for-information-public-access-to-peer-reviewed-scholarly-publications-data-and-code > > Please note that OSTP is also considering requiring that code and data > necessary to reproduce results be made available as well, so feel free > to weigh in on that as well. > > Thank you, > > Matt Thank you for sharing. Can I ask two questions (not including this one)? 1) I am not familiar with OSTP: is there a template one can use to respond to the RFI? 2) Who can submit a comment? Anyone living in US? Any US citizen? I am asking in order to assess where to disseminate this. Regards, Marco Caminati From Isaac.Oscar.Gariano at ecs.vuw.ac.nz Mon Mar 2 16:59:13 2020 From: Isaac.Oscar.Gariano at ecs.vuw.ac.nz (Isaac) Date: Mon, 2 Mar 2020 21:59:13 +0000 Subject: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? In-Reply-To: References: , Message-ID: Dear Ben, Thank you! I've looked at the source code for the type-script type-checker (https://github.com/microsoft/TypeScript/blob/master/src/compiler/checker.ts), and it appears to be doing two things: * If the type-aliases are the same (i.e. if computing A sub-type of A) than it infers the variance of the type-paramaters of A (by using the getAliasVariances function), and then checks each Ti and Yi against this variance information. * If a sub-typing check involving the same type-alias recurs 5 times, it simply returns true (this is done by the isDeeplyNestedType function). So my ExplodingList example is excepted as it uses variance to infer that ExplodingList is a sub-type of ExplodingList if T is a sub-type of Y; and so it deduces that ExplodingList is a sub-type of ExplodingList> since by the con-inductive hypothesis, IntegerExplosion is a sub-type of ExplodingList. However this variance check dosn't occur if the type-aliases are different but just happen to have the same structure, e.g. type ExplodingList1 = { head: T, explode: ExplodingList1> }; type ExplodingList2 = { head: T, explode: ExplodingList2>}; When checking if ExplodingList1 is a sub-type of ExplodingList2 it simply hits the recursion limit of 5 and returns true. Obviously this recursion depth limit is unsound, consider the following two examples: type A = { head: T1, tail: A } type B = { head: T1, tail: B } function sub1(x: A): B { return x } type N = { head: Y } function sub2(x: N>>>>): N>>>> { return x } Both are accepted as well-typed by type-script. What I find interesting is the second example does not actually involve recursive types, yet it still hits the limit; this is because type-script simply checks whether it's already in a sub-typing check of the same alias-name (in this case N) even if these uses of "N" are unrelated. I can think of refining this test in two ways to reject the above examples, but I'm sure there are many more that it will still accept: * Make the recursion depth limit be relative to the number of type-parameters (so the cycling trick of my first example would be rejected) * Only count applications of the type-alias that came from the same place in the program (so any recursive applications of A<...> sub-type of B<...> where the A and B applications came from the "tail" field will count towards the recursion limit, but N<...> sub-type of N<...> will not if the "N"s came from different locations in the type signature of "sub2"). The idea is that the recursion limit will only hit if it is actually the result of a type-alias being recursive (and not just having a nested application like my N example). Let me know if my understanding of type-script is incorrect! ? Isaac Oscar Gariano? ________________________________ From: Types-list on behalf of Benjamin Lichtman Sent: 19 February 2020 11:14 AM To: Martin Abadi Cc: types-list at lists.seas.upenn.edu Subject: Re: [TYPES] Has anyone seen a language with sub-typing and recursive parametric type aliases? [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] Hello all, TypeScript team member here. This example actually is supported by TypeScript. See here for a demonstration. Best, Ben On Tue, Feb 18, 2020 at 1:14 PM Martin Abadi wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Hello, > > TypeScript has parameterized recursive > definitions and structural subtyping. The language has evolved over time, > and I am afraid I have not looked at the current definitions, but, early > on, the designers were aware of some of the algorithmic difficulties > that Fran?ois Pottier mentions and introduced restrictions to avoid them. I > don't know whether the examples of interest to Isaac Oscar Gariano would be > expressible. > > Regards, > Martin > > > On Tue, Feb 18, 2020 at 10:36 AM Fran?ois Pottier < > francois.pottier at inria.fr> > wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > > > Hello, > > > > If I am not mistaken, already deciding the *equivalence* of two types > > in the presence of recursive and parameterized type abbreviations is > > extremely costly: Marvin Solomon proved it equivalent to the DPDA > > equivalence problem ("Type definitions with parameters", POPL'78). > > > > -- > > Fran?ois Pottier > > francois.pottier at inria.fr > > http://cambium.inria.fr/~fpottier/ > > > From wadler at inf.ed.ac.uk Thu Mar 26 14:10:27 2020 From: wadler at inf.ed.ac.uk (Philip Wadler) Date: Thu, 26 Mar 2020 15:10:27 -0300 Subject: [TYPES] A question about the literature on monomorphisation Message-ID: Dear Types, We have a question about the literature on monomorphisation. We will give an example in terms generics for Java, but similar features occur in many other languages (C++ templates, C#, Rust, and so on) and we are interested in all of these. Java generics are implemented by erasure, but let's consider what is required for implementation by monomorphisation. In Java, each of classes, interfaces, and methods may take generic parameters. Monomorphisation replaces each generic class, interface, or method by one instance for each actual generic parameter that occurs in the program. An interface that is generic may itself contain generic methods. In that case, monomorphisation needs to generate one instance of the interface for each type at which it is instantiated, and then one instance of the method within the interface for each type at which *it* is instantiated. For example, consider the following Java interface: interface List { public String toString() public List map(f Function(A,B)) } and this code: List ints List bools Function square Function positive String string1 = ints.toString() List ints2 = ints.map(square) List bools2 = int.map(positive) String string2 = bools.ToString() If there are no other invocations of toString() or map(), the List interface would have two instances, the first requiring two instances of the map method and the second requiring none: interface ListInteger { public String toString() public ListInteger mapInteger(f FunctionIntegerInteger) public ListBoolean mapBoolean(f FunctionIntegerBoolean) } interface ListBoolean { public String toString() } The bookkeeping required to formalise monomorphisation of instances and methods is not too difficult, but not trivial---it took us several tries to formalise it for a different language. The question is, where has this been described in the literature? Below is a list of the references we have found on monomorphisation, but none of them seems to cover the specific problem of monomorphisation of generic methods within generic interfaces, as described above, even though that combination occurs in several languages. Have we missed any relevant literature? Thank you for your help. Raymond Hu Wen Kokke Julien Lange Bernardo Toninho Philip Wadler Nobuko Yoshida References: From ML to Ada: Strongly-Typed Language Interoperability via Source Translation. Andrew P. Tolmach and Dino Oliva, JFP 1998. Design and Implementation of Generics for the .NET Common Language Runtime. Andrew Kennedy and Don Syme, PLDI 2001. Formalization of generics for the .NET common language runtime. Dachuan Yu, Andrew Kennedy, and Don Syme, POPL 2004. A Semantic Analysis of C++ Templates. Jeremy Siek and Walid Taha, ECOOP 2006. Whole-program compilation in MLton. Stephen Weeks. ML 2006. (Slides of invited talk.) Encoding Monomorphic and Polymorphic Types. Jasmin Christian Blanchette, Sascha B?hme, Andrei Popescu, and Nicholas Smallbone, Logical Methods in Computer Science 2016. Safe Low-level Code Generation in Coq Using Monomorphization and Monadification. Akira Tanaka, Reynald Affeldt, and Jacques Garrigue. JIP 2018. Monomorphisation for Rust (commented code) https://doc.rust-lang.org/nightly/nightly-rustc/rustc_mir/monomorphize/collector/index.html Monomorphisation for Mlton (commented code) https://rustc-dev-guide.rust-lang.org/backend/monomorph.html . \ Philip Wadler, Professor of Theoretical Computer Science, . /\ School of Informatics, University of Edinburgh . / \ and Senior Research Fellow, IOHK . http://homepages.inf.ed.ac.uk/wadler/ -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: not available URL: From kthielen at gmail.com Thu Mar 26 17:21:54 2020 From: kthielen at gmail.com (Kalani Thielen) Date: Thu, 26 Mar 2020 17:21:54 -0400 Subject: [TYPES] A question about the literature on monomorphisation In-Reply-To: References: Message-ID: <5816D400-53DE-4F93-97CA-C5C35D147137@gmail.com> Mark Jones?s dissertation on qualified types has a useful section on monomorphisation near the end. In that context it was to eliminate type constraints (e.g. type classes) by rewriting, so resolving overloading and monomorphising at the same time. Polymorphic definitions without qualifications and nested definitions can be handled the same way if you don?t mind generating hidden constraints/classes. This is the route we took at Morgan Stanley for a PL we use internally. HTH > On Mar 26, 2020, at 2:10 PM, Philip Wadler wrote: > > The bookkeeping required to formalise monomorphisation of instances > and methods is not too difficult, but not trivial---it took us several > tries to formalise it for a different language. > > The question is, where has this been described in the literature? > Below is a list of the references we have found on monomorphisation, > but none of them seems to cover the specific problem of > monomorphisation of generic methods within generic interfaces, as > described above, even though that combination occurs in several > languages. Have we missed any relevant literature? From wadler at inf.ed.ac.uk Mon Mar 30 10:36:58 2020 From: wadler at inf.ed.ac.uk (Philip Wadler) Date: Mon, 30 Mar 2020 11:36:58 -0300 Subject: [TYPES] A question about the literature on monomorphisation In-Reply-To: <5816D400-53DE-4F93-97CA-C5C35D147137@gmail.com> References: <5816D400-53DE-4F93-97CA-C5C35D147137@gmail.com> Message-ID: Thanks. I downloaded a copy of Mark's thesis from here: https://www.cs.ox.ac.uk/files/3432/PRG106.pdf It refers to the "monomorphism restriction" a few times, but I cannot find a reference to "monomorphisation". Can you say which sections you had in mind? If you have an example of monomorphisation for the Morgan Stanley PL you mention that is similar to the example in my original question, I would love to see it. Is your PL described publicly anywhere? Cheers, -- P . \ Philip Wadler, Professor of Theoretical Computer Science, . /\ School of Informatics, University of Edinburgh . / \ and Senior Research Fellow, IOHK . http://homepages.inf.ed.ac.uk/wadler/ On Thu, 26 Mar 2020 at 18:22, Kalani Thielen wrote: > Mark Jones?s dissertation on qualified types has a useful section on > monomorphisation near the end. In that context it was to eliminate type > constraints (e.g. type classes) by rewriting, so resolving overloading and > monomorphising at the same time. Polymorphic definitions without > qualifications and nested definitions can be handled the same way if you > don?t mind generating hidden constraints/classes. This is the route we > took at Morgan Stanley for a PL we use internally. > > HTH > > > On Mar 26, 2020, at 2:10 PM, Philip Wadler wrote: > > The bookkeeping required to formalise monomorphisation of instances > and methods is not too difficult, but not trivial---it took us several > tries to formalise it for a different language. > > The question is, where has this been described in the literature? > Below is a list of the references we have found on monomorphisation, > but none of them seems to cover the specific problem of > monomorphisation of generic methods within generic interfaces, as > described above, even though that combination occurs in several > languages. Have we missed any relevant literature? > > > -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: not available URL: From kthielen at gmail.com Mon Mar 30 14:30:04 2020 From: kthielen at gmail.com (Kalani Thielen) Date: Mon, 30 Mar 2020 14:30:04 -0400 Subject: [TYPES] A question about the literature on monomorphisation In-Reply-To: References: <5816D400-53DE-4F93-97CA-C5C35D147137@gmail.com> Message-ID: I think that the print copy has this section that isn?t anywhere online. I can?t verify at the moment though, because I gave my print copy to a friend. Our PL at Morgan Stanley is https://github.com/Morgan-Stanley/hobbes/ , basically an eager variant of Haskell with structural types (and a few other features useful in our area). We monomorphise code to avoid penalties for polymorphism/overloading (we are very sensitive to pre-trade latency). The approach here (and IIRC the same way Mark Jones described it) is to say that only monomorphic constraints can be satisfied, and that satisfied constraints are eliminated by rewriting qualified expressions into equivalent expressions without qualification (e.g. show(42) :: Show int => [char] ? rewrites to ?> showInt(42) :: [char]). That winds up being an API for user code to define new ways to resolve constraints (e.g. network RPC, where connections are made and protocols negotiated at ?compile time?). It looks a little different than your example, and there?s more going on than just monomorphisation, but the kind of nested polymorphism you describe gets ?flattened out? and we generate hidden type classes (e.g. List { List map(A->B) } would be in a subclass from List { String toString() } to add the additional type variable). On the back end we wind up with a lot of tiny type classes with long parameter lists. HTH, sorry if I?ve taken a tangent. :) > On Mar 30, 2020, at 10:36 AM, Philip Wadler wrote: > > Thanks. I downloaded a copy of Mark's thesis from here: > https://www.cs.ox.ac.uk/files/3432/PRG106.pdf > > It refers to the "monomorphism restriction" a few times, but I cannot find a reference to "monomorphisation". Can you say which sections you had in mind? > > If you have an example of monomorphisation for the Morgan Stanley PL you mention that is similar to the example in my original question, I would love to see it. Is your PL described publicly anywhere? > > Cheers, -- P > > . \ Philip Wadler, Professor of Theoretical Computer Science, > . /\ School of Informatics, University of Edinburgh > . / \ and Senior Research Fellow, IOHK > . http://homepages.inf.ed.ac.uk/wadler/ > > > > On Thu, 26 Mar 2020 at 18:22, Kalani Thielen > wrote: > Mark Jones?s dissertation on qualified types has a useful section on monomorphisation near the end. In that context it was to eliminate type constraints (e.g. type classes) by rewriting, so resolving overloading and monomorphising at the same time. Polymorphic definitions without qualifications and nested definitions can be handled the same way if you don?t mind generating hidden constraints/classes. This is the route we took at Morgan Stanley for a PL we use internally. > > HTH > > >> On Mar 26, 2020, at 2:10 PM, Philip Wadler > wrote: >> >> The bookkeeping required to formalise monomorphisation of instances >> and methods is not too difficult, but not trivial---it took us several >> tries to formalise it for a different language. >> >> The question is, where has this been described in the literature? >> Below is a list of the references we have found on monomorphisation, >> but none of them seems to cover the specific problem of >> monomorphisation of generic methods within generic interfaces, as >> described above, even though that combination occurs in several >> languages. Have we missed any relevant literature? > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. From tadeusz.litak at gmail.com Tue Apr 14 00:21:26 2020 From: tadeusz.litak at gmail.com (Tadeusz Litak) Date: Tue, 14 Apr 2020 06:21:26 +0200 Subject: [TYPES] Guide to best practices for virtual conferences In-Reply-To: <041ED45D-D693-4807-AA6B-45B3FF484F3D@cis.upenn.edu> References: <041ED45D-D693-4807-AA6B-45B3FF484F3D@cis.upenn.edu> Message-ID: <9b5f2525-9edf-9844-db0d-82289e304503@gmail.com> Dear Benjamin, dear all, apologies for replying via the mailing list (I obviously chose Types rather than Types-announce) instead of heading towards the live Google Doc you mentioned. However, I believe that the importance of the issue I want to raise goes beyond the task force report in question, especially given that all meetings, workshops and conferences are rapidly transferred into online format. Namely, the report appears to heavily promote Zoom as a go-to tool for various meeting formats without even mentioning in passing, much less discussing in detail growing questions about its darker side, in particular concerning privacy and security. By now Zoom has been banned by numerous educational institutions (examples: schools in New York, Berkeley or Singapore, but also, e.g., Leiden University) and various other organizations, companies and governments: https://en.wikipedia.org/wiki/Zoom_Video_Communications#Bans Given everything that I have read so far, I believe these bans are justified. To stop this backlash, Zoom has announced 90-day feature freeze focusing solely on privacy and security, but also more or less admitting in passing that in all nine years of company's existence, they have never taken those questions seriously (and this is a favourable interpretation; I am personally leaning towards a less charitable one). It remains to be seen how independent security experts are going to evaluate the outcome of these announced three-month efforts. But as things stand now, especially given? the inquiry launched by the? New York State Attorney General, uncritically recommending the tool seems simply irresponsible, or at the very least premature. Inasmuch as I can see briefly scanning through the task force report, it does not answer basic questions in this regard: * What is the situation of potential attendees and especially speakers who are unwilling to use Zoom or perhaps even in some cases banned from using it? * Are conference organizers encouraged to investigate if their potential audience agrees with the use of Zoom (or any other specific software chosen for the event)? * Is every effort made to ensure that everybody involved---not just the organizers, but also speakers and participants--- is aware of security and privacy risks involved and ways of minimizing those risks? Just telling the hosts to be on the constant lookout for "Zoombombing", as the report suggests, seems fairly minimalistic to me (apart from the fact that if "Zoombombing" is possible at all, then it means that security settings of a given meeting are rather lax) * Have alternative solutions (proprietary and open-source, free or commercial) been sufficiently investigated and evaluated? Are people in the community even aware? of the full spectrum of options? The last point brings us to another issue. Zoom is not only closed source and proprietary, but its free licence entails certain limitations from organizers' point of view, especially the restriction of group meetings to 40 minutes. In effect, organizers of any event would either need to purchase a suitable license or insist that their institution provides one. If a license needs to be purchased by the organizers at all, then there are commercial tools which handle poor connections no worse or better than Zoom does, providing crisp video and clear voice, but unlike Zoom offer unmanipulated in-browser participation, without trying to force full-blown installation of the full desktop client, for example. I am still in the process of testing various solutions, but I have had positive experiences with, e.g., EyesOn. And there is always Jitsi, which is free and open source; unfortunately, in terms of handling video calls with multiple participants over shaky connections, it is not quite up there with the likes of EyesOn. Many educational institutions already have licenses for other tried and broadly used software, most likely MS Teams. There is yet another related, but much broader and more fundamental problem which the task force report does not seem to address. Namely, the unfolding disaster should lead to a badly overdue discussion of the future of conference-based model of dissemination of results in TCS. It has long been seen as a historical pathology, not only by the vast majority of other disciplines, but also by numerous leading researchers, much more accomplished than myself. It has been distorting normal refereeing and publication process, encouraging and rewarding publication of "extended abstracts" instead of properly vetted and written full papers. There was also obvious criticism regarding carbon footprint and especially concerning participation and publication costs, this last issue being intensely discussed on this very mailing list. Now, while the carbon footprint is no longer a concern for online events, the issue of "extended abstracts" and viability of this entire publication model is not going anywhere. And we have now a whole new bunch of problems discussed above (as well as numerous other ones discussed in the task force report), while the usual argument about benefits of face-to-face meetings loses most of its force. Why are we keeping up the whole charade? These days, it is possible to augment journal papers with sets of slides or video presentations, if there is a need for that. Grant money and funding that used to go towards either organization of conferences or participation in such events these days would be much better spent on converting them into open-access journals (and as has been pointed out on this very mailing list, ACM's Gold Open Access does appear overpriced). Ditto for the time and energy of everybody involved in the organization of said meetings; and as the task force report confirms, they will require no less preparation than flesh-and-blood meetings. Sure, with journal publications there are issues regarding turnover, competitiveness, timely feedback; the extended abstract model also has the advantage of enforcing conciseness. But I do not think that these issues cannot be overcome when converting at least some of existing conference series into online journals, preferably open access. Efforts in this direction seem more worthwhile to me than trying keep everything going in the form of Zoom webinars. There is also the potential of coming up with new publication models. My long-standing pet peeve, for example, has been the question why most proof assistants do not have anything remotely comparable to Isabelle's Archive of Formal Proofs. If we, however, do insist on making online conferences the basic dissemination model, I think that questions from the first part of my overlong email should be addressed. Kind regards, t. On 13.04.20 14:28, Benjamin C. Pierce wrote: > [ The Types Forum (announcements only), > http://lists.seas.upenn.edu/mailman/listinfo/types-announce ] > > > *[TL;DR: Please help circulate a new guide for organizers of virtual conferences.]* > > Dear colleagues, > > The Association for Computing Machinery recently chartered a Presidential Task Force to gather and disseminate > guidance on best practices for virtual conferences, aimed at the many conference organizers moving their events online > right now. > > The task force report, *Virtual Conferences: A Guide to Best Practices , *is > now available on the ACM web site ? It offers a comprehensive survey of > issues, organizational strategies, and technology platforms for successful virtual meetings. > > We hope that you and others in your field will find this report useful. If you do, we would love to hear about it! And > naturally if you have any suggestions for improvement, we would love to hear those too; the PDF document linked above > includes a pointer to a live Google Doc where you can leave suggestions and comments if you like. If you have recently > organized a virtual conference or are organizing one now, we would especially like to include your experiences (how > you organized it, how it went, what people thought, a summary of any post-conference survey results, your advice for > future conferences, etc.) and add it (or better yet a pointer to it) to the appendix that we?ve provided for such > experience reports. > > Finally, can you please help us make sure this guide reaches the people that need it by forwarding this announcement > within your networks (especially, of course, to current conference organizers)? > > Many thanks! > > ? ? ?Benjamin Pierce > > ?on behalf of the entire task force: > > Crista Videira Lopes , University of California, Irvine, USA (Task Force Co-chair) > Jeanna Matthews , Clarkson University, USA (Task Force Co-chair, member of ACM > Council, Former SGB Chair) > Benjamin Pierce , University of Pennsylvania, USA (Task Force Executive Editor, > SIGPLAN Vice Chair, chair of SIGPLAN ad hoc committee on climate change ) > > Blair MacIntyre , Georgia Tech, USA (Chaired IEEE VR 2020) > Gary Olson , University of California, Irvine, USA (Former SIGCHI Treasurer; Chair of CSCW > Steering Committee, chaired CHI, CSCW, DIS, and many non-ACM conferences) > Rob Lindeman , University of > Canterbury, NZ (Chaired IEEE VR 2010) > Francois Guimbretiere , Cornell University, USA (Chaired UIST 2019) > Srinivasan Keshav , University of Cambridge, UK (Former > SIGCOMM Chair) > > Ex-officio members: > Vicki Hanson (ACM CEO, Former ACM President) > Pat Ryan (ACM COO) > Donna Cappo (ACM Director of SIG Services) > From bcpierce at cis.upenn.edu Tue Apr 14 11:05:41 2020 From: bcpierce at cis.upenn.edu (Benjamin C. Pierce) Date: Tue, 14 Apr 2020 11:05:41 -0400 Subject: [TYPES] Guide to best practices for virtual conferences In-Reply-To: <9b5f2525-9edf-9844-db0d-82289e304503@gmail.com> References: <041ED45D-D693-4807-AA6B-45B3FF484F3D@cis.upenn.edu> <9b5f2525-9edf-9844-db0d-82289e304503@gmail.com> Message-ID: <22AE3B7B-B008-4336-A882-A6AA48124E17@cis.upenn.edu> Thank you for these comments, Tadeusz! The discussion around Zoom has been developing quite fast, and the task force is considering whether a change in recommendations is appropriate. For the moment, we will include links to discussions of its security and privacy issues. As to the larger issue of whether we should be putting energy into virtualizing conferences at all (vs. throwing away the whole conference model), there is obviously room for a larger discussion here, and I hope others will chime in. My own sense is that the conference model is already in the process of changing in fundamental ways, as the short-term pressure of COVID-19 amplifies the long-term pressure of climate change (which we were already beginning to take seriously). Exactly what the new steady state will be in a few years or a decade is hard to predict, but I think it?s likely to include both a stronger role for asynchronous modes of communication (things like journals but also personal web sites, Arxiv, YouTube channels, etc.) as well as forums for synchronous communication that support a range of interaction styles ? structured and unstructured, planned and serendipitous, large group, small group, and 1-1, etc. We may even call the latter ?conferences.? :-) Best, - Benjamin > On Apr 14, 2020, at 12:21 AM, Tadeusz Litak wrote: > > Dear Benjamin, dear all, > > apologies for replying via the mailing list (I obviously chose Types rather than Types-announce) instead of heading towards the live Google Doc you mentioned. However, I believe that the importance of the issue I want to raise goes beyond the task force report in question, especially given that all meetings, workshops and conferences are rapidly transferred into online format. > > Namely, the report appears to heavily promote Zoom as a go-to tool for various meeting formats without even mentioning in passing, much less discussing in detail growing questions about its darker side, in particular concerning privacy and security. By now Zoom has been banned by numerous educational institutions (examples: schools in New York, Berkeley or Singapore, but also, e.g., Leiden University) and various other organizations, companies and governments: > > https://en.wikipedia.org/wiki/Zoom_Video_Communications#Bans > > Given everything that I have read so far, I believe these bans are justified. To stop this backlash, Zoom has announced 90-day feature freeze focusing solely on privacy and security, but also more or less admitting in passing that in all nine years of company's existence, they have never taken those questions seriously (and this is a favourable interpretation; I am personally leaning towards a less charitable one). It remains to be seen how independent security experts are going to evaluate the outcome of these announced three-month efforts. But as things stand now, especially given the inquiry launched by the New York State Attorney General, uncritically recommending the tool seems simply irresponsible, or at the very least premature. > > Inasmuch as I can see briefly scanning through the task force report, it does not answer basic questions in this regard: > > What is the situation of potential attendees and especially speakers who are unwilling to use Zoom or perhaps even in some cases banned from using it? > Are conference organizers encouraged to investigate if their potential audience agrees with the use of Zoom (or any other specific software chosen for the event)? > Is every effort made to ensure that everybody involved---not just the organizers, but also speakers and participants--- is aware of security and privacy risks involved and ways of minimizing those risks? Just telling the hosts to be on the constant lookout for "Zoombombing", as the report suggests, seems fairly minimalistic to me (apart from the fact that if "Zoombombing" is possible at all, then it means that security settings of a given meeting are rather lax) > Have alternative solutions (proprietary and open-source, free or commercial) been sufficiently investigated and evaluated? Are people in the community even aware of the full spectrum of options? > The last point brings us to another issue. Zoom is not only closed source and proprietary, but its free licence entails certain limitations from organizers' point of view, especially the restriction of group meetings to 40 minutes. In effect, organizers of any event would either need to purchase a suitable license or insist that their institution provides one. > > If a license needs to be purchased by the organizers at all, then there are commercial tools which handle poor connections no worse or better than Zoom does, providing crisp video and clear voice, but unlike Zoom offer unmanipulated in-browser participation, without trying to force full-blown installation of the full desktop client, for example. I am still in the process of testing various solutions, but I have had positive experiences with, e.g., EyesOn. And there is always Jitsi, which is free and open source; unfortunately, in terms of handling video calls with multiple participants over shaky connections, it is not quite up there with the likes of EyesOn. Many educational institutions already have licenses for other tried and broadly used software, most likely MS Teams. > There is yet another related, but much broader and more fundamental problem which the task force report does not seem to address. Namely, the unfolding disaster should lead to a badly overdue discussion of the future of conference-based model of dissemination of results in TCS. It has long been seen as a historical pathology, not only by the vast majority of other disciplines, but also by numerous leading researchers, much more accomplished than myself. It has been distorting normal refereeing and publication process, encouraging and rewarding publication of "extended abstracts" instead of properly vetted and written full papers. There was also obvious criticism regarding carbon footprint and especially concerning participation and publication costs, this last issue being intensely discussed on this very mailing list. Now, while the carbon footprint is no longer a concern for online events, the issue of "extended abstracts" and viability of this entire publication model is not going anywhere. And we have now a whole new bunch of problems discussed above (as well as numerous other ones discussed in the task force report), while the usual argument about benefits of face-to-face meetings loses most of its force. > > Why are we keeping up the whole charade? These days, it is possible to augment journal papers with sets of slides or video presentations, if there is a need for that. Grant money and funding that used to go towards either organization of conferences or participation in such events these days would be much better spent on converting them into open-access journals (and as has been pointed out on this very mailing list, ACM's Gold Open Access does appear overpriced). Ditto for the time and energy of everybody involved in the organization of said meetings; and as the task force report confirms, they will require no less preparation than flesh-and-blood meetings. Sure, with journal publications there are issues regarding turnover, competitiveness, timely feedback; the extended abstract model also has the advantage of enforcing conciseness. But I do not think that these issues cannot be overcome when converting at least some of existing conference series into online journals, preferably open access. Efforts in this direction seem more worthwhile to me than trying keep everything going in the form of Zoom webinars. > > There is also the potential of coming up with new publication models. My long-standing pet peeve, for example, has been the question why most proof assistants do not have anything remotely comparable to Isabelle's Archive of Formal Proofs. > > If we, however, do insist on making online conferences the basic dissemination model, I think that questions from the first part of my overlong email should be addressed. > > Kind regards, > > t. > > > > On 13.04.20 14:28, Benjamin C. Pierce wrote: >> [ The Types Forum (announcements only), >> http://lists.seas.upenn.edu/mailman/listinfo/types-announce ] >> >> >> >> [TL;DR: Please help circulate a new guide for organizers of virtual conferences.] >> >> Dear colleagues, >> >> The Association for Computing Machinery recently chartered a Presidential Task Force to gather and disseminate guidance on best practices for virtual conferences, aimed at the many conference organizers moving their events online right now. >> >> The task force report, Virtual Conferences: A Guide to Best Practices , is now available on the ACM web site It offers a comprehensive survey of issues, organizational strategies, and technology platforms for successful virtual meetings. >> >> We hope that you and others in your field will find this report useful. If you do, we would love to hear about it! And naturally if you have any suggestions for improvement, we would love to hear those too; the PDF document linked above includes a pointer to a live Google Doc where you can leave suggestions and comments if you like. If you have recently organized a virtual conference or are organizing one now, we would especially like to include your experiences (how you organized it, how it went, what people thought, a summary of any post-conference survey results, your advice for future conferences, etc.) and add it (or better yet a pointer to it) to the appendix that we?ve provided for such experience reports. >> >> Finally, can you please help us make sure this guide reaches the people that need it by forwarding this announcement within your networks (especially, of course, to current conference organizers)? >> >> Many thanks! >> >> Benjamin Pierce >> >> ?on behalf of the entire task force: >> >> Crista Videira Lopes , University of California, Irvine, USA (Task Force Co-chair) >> Jeanna Matthews , Clarkson University, USA (Task Force Co-chair, member of ACM Council, Former SGB Chair) >> Benjamin Pierce , University of Pennsylvania, USA (Task Force Executive Editor, SIGPLAN Vice Chair, chair of SIGPLAN ad hoc committee on climate change ) >> >> Blair MacIntyre , Georgia Tech, USA (Chaired IEEE VR 2020) >> Gary Olson , University of California, Irvine, USA (Former SIGCHI Treasurer; Chair of CSCW Steering Committee, chaired CHI, CSCW, DIS, and many non-ACM conferences) >> Rob Lindeman , University of Canterbury, NZ (Chaired IEEE VR 2010) >> Francois Guimbretiere , Cornell University, USA (Chaired UIST 2019) >> Srinivasan Keshav , University of Cambridge, UK (Former SIGCOMM Chair) >> >> Ex-officio members: >> Vicki Hanson (ACM CEO, Former ACM President) >> Pat Ryan (ACM COO) >> Donna Cappo (ACM Director of SIG Services) >> > From monnier at iro.umontreal.ca Tue Apr 14 11:30:22 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Tue, 14 Apr 2020 11:30:22 -0400 Subject: [TYPES] Guide to best practices for virtual conferences In-Reply-To: <9b5f2525-9edf-9844-db0d-82289e304503@gmail.com> (Tadeusz Litak's message of "Tue, 14 Apr 2020 06:21:26 +0200") References: <041ED45D-D693-4807-AA6B-45B3FF484F3D@cis.upenn.edu> <9b5f2525-9edf-9844-db0d-82289e304503@gmail.com> Message-ID: > Namely, the report appears to heavily promote Zoom as a go-to tool for > various meeting formats without even mentioning in passing, much less > discussing in detail growing questions about its darker side, in particular > concerning privacy and security. Agreed. I'm particularly worried about imposing the use of proprietary software on the conference's participants. Stefan From jonathan.aldrich at cs.cmu.edu Wed Apr 29 19:48:29 2020 From: jonathan.aldrich at cs.cmu.edu (Jonathan Aldrich) Date: Wed, 29 Apr 2020 19:48:29 -0400 Subject: [TYPES] effect vs. coeffect expressiveness Message-ID: Dear Types, I am curious about the relative expressiveness of effects and coeffects. Has this been studied? The cleanest distinction I've seen is that effects capture the impact a program has on its environment, i.e. what it produces. Coeffects capture the requirements that a program puts on its environment: what it consumes. This is discussed, for example, in Gaboardi et al's ICFP 2016 paper, "Combining Effects and Coeffects via Grading" (and elsewhere). There is some useful intuition in this distinction, and it describes the different structure of checking rules in effect and coeffect systems. However, I don't find this distinction very helpful in thinking about expressiveness. It seems like many examples can be expressed in either an effect or a coeffect system. For example, an exception is a classic example of an effect (e.g. in the paper mentioned above, and many others). However, it seems to me that exceptions can also be modeled as coeffects: code that might throw an exception requires the caller to pass a handler for that exception to it--or perhaps an abstract "permission" to throw that exception. So in what sense are exceptions an effect, rather than a coeffect? Is this true of all the kinds of things that are typically expressed with effects and coeffects--that they could just as easily be expressed in the other style? If so, what are the benefits of one style vs. the other? Or are there examples that can only be expressed in one style--or for which expression in the other style is much more awkward? Perhaps these questions have been written about, but I haven't been able to find it. I would love to get some pointers. I am particularly interested in a practical explanation of the differences in expressiveness, or theoretical results that have a direct and explicit relationship to practice--with practical examples in either case. (co-)Monads and/or anything categorical are not a very helpful starting point for me, but effect systems and/or linear types are (I've done research on both). I am also more interested in the descriptive view of effects/coeffects (in the sense of Filinski, ICFP'11) than the prescriptive view. Thanks, Jonathan From breuvart at lipn.univ-paris13.fr Thu Apr 30 04:06:32 2020 From: breuvart at lipn.univ-paris13.fr (Flavien Breuvart) Date: Thu, 30 Apr 2020 10:06:32 +0200 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: <82231950-c440-fb97-6089-efde0ab83b20@lipn.univ-paris13.fr> Hi Jonathan, There is not much more written as there is a lot that we do not understand here. I will present my point of view on the subject (this is an invitation for other point of view to express themselves ;). First of all, I may have been attracted by the production/consumption distinction, I do not use it anymore. It is miss-leading and not that usefull as you said. One thing is certain : monads are way more natural in call-by-value, and comonads are way more natural in call-by-name. I have to stop Haskeller complaints as an aparte : most monad in Haskell are intrincically CbV as the argument is evaluated to some extent by the bind, in addition Haskell is call-by-need, which is slightly different from this point of view. The real distinction I use on effects/coeffects is related to this statement : *effects describe what happen when the piece of code is evaluated*, while *coeffects describe what append when it is manipulated *(copied/erased/distributed). Following this idea, one can see that most coeffects can be somehow encapsulated into effects (basically by looking at what happened to the other copies of the codes); and worst, all applied coeffect on the literature are internally encoded this way since code manipulations are not accessible by programmers. But I believe that there is more to be discovered and used from coeffect, especially for writing concurrent programs or performing code analyses. I was a bit expeditious,? I can develop some of the points if you are interested. I hope you enjoyed those random thoughts, and I invite other to do the same. This subject is very interesting, but too fuzzy for it to be formalized yet :) Cheers, Flavien (the one from Gaboardi et al's ICFP paper) Le 30/04/2020 ? 01:48, Jonathan Aldrich a ?crit?: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Types, > > I am curious about the relative expressiveness of effects and > coeffects. Has this been studied? > > The cleanest distinction I've seen is that effects capture the impact > a program has on its environment, i.e. what it produces. Coeffects > capture the requirements that a program puts on its environment: what > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > 2016 paper, "Combining Effects and Coeffects via Grading" (and > elsewhere). > > There is some useful intuition in this distinction, and it describes > the different structure of checking rules in effect and coeffect > systems. However, I don't find this distinction very helpful in > thinking about expressiveness. It seems like many examples can be > expressed in either an effect or a coeffect system. For example, an > exception is a classic example of an effect (e.g. in the paper > mentioned above, and many others). However, it seems to me that > exceptions can also be modeled as coeffects: code that might throw an > exception requires the caller to pass a handler for that exception to > it--or perhaps an abstract "permission" to throw that exception. So > in what sense are exceptions an effect, rather than a coeffect? > > Is this true of all the kinds of things that are typically expressed > with effects and coeffects--that they could just as easily be > expressed in the other style? If so, what are the benefits of one > style vs. the other? Or are there examples that can only be expressed > in one style--or for which expression in the other style is much more > awkward? > > Perhaps these questions have been written about, but I haven't been > able to find it. I would love to get some pointers. I am > particularly interested in a practical explanation of the differences > in expressiveness, or theoretical results that have a direct and > explicit relationship to practice--with practical examples in either > case. (co-)Monads and/or anything categorical are not a very helpful > starting point for me, but effect systems and/or linear types are > (I've done research on both). I am also more interested in the > descriptive view of effects/coeffects (in the sense of Filinski, > ICFP'11) than the prescriptive view. > > Thanks, > > Jonathan From dom.orchard at gmail.com Thu Apr 30 04:40:25 2020 From: dom.orchard at gmail.com (Dominic Orchard) Date: Thu, 30 Apr 2020 09:40:25 +0100 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: Dear Jonathan, This is a great question and something that I think still lacks a thorough comprehensive treatment in the literature, but certainly there are hints about this in various places (and I have been slowly feeling my way towards different versions of this question, including some work in progress right now!) This is certainly something that we are trying to explore in the Granule project (https://granule-project.github.io/, and at ICFP 2019) where we have a fully functioning modern functional programming language which can work with various different effects and coeffects at the same time. We are brewing a few things comparing certain analyses performed as effects vs coeffects (because, as you say, it seems that some things can be done both ways, so why choose one over the other?). It's probably worth pointing out that, whilst there is quite a bit of recent work with coeffects in a linear type theory (like Granule and the ICFP 2016 paper you mention), coeffects need not be tided to a linear type theory (indeed, the first two papers that Tomas, Alan, and I wrote on this were in a Cartesian setting (ICALP 2013 and ICFP 2014)). My general, informal view is that (broadly) monadic effects capture a kind of 'forwards' analysis, with information flowing from inputs to outputs, whereas comonadic coeffects capture an 'backwards'-style analysis, flowing back through the inputs. How you structure the primitives for working with either will affect how easy it is to capture certain properties (and the CBN/CBV story then becomes important as well, see the Cicek et al. paper mentioned below). Here are a few references that spring to mind that go a little bit towards your question: * Tomas Petricek and I reflected on coeffect vs effect expressivity a bit in Section 8 of this paper ( https://www.cs.kent.ac.uk/people/staff/dao7/publ/haskell14-effects.pdf) where we give a brief comparison between expressing implicit parameters as effects vs coeffects (reader graded monad vs product graded comonad) and what the difference is (we point out that Haskell's implicit parameters feature are a coeffect rather than an effect, perhaps the most "mainstream" example of explicit coeffects in the wild!). * This short paper by Ezgi Cicek, Marco Gaboardi, and Deepak Garg looks at this question in the context of cost analysis ( https://lipn.univ-paris13.fr/DICE2016/Abstracts/paper_10.pdf). * I know that you said a (co)monad treatment is not helpful to you, but in case it helps others, I have this unpublished article from the end of my PhD "Should I use a Monad or Comonad?" which at least explains some overlap ( https://www.cs.kent.ac.uk/people/staff/dao7/drafts/monad-or-comonad-orchard11-draft.pdf) and tries to reconcile some tensions in the literature. The paper is meant to be a fairly self-contained. * The ICFP 2018 by Andrew Hirsch and Ross Tate ( http://www.cs.cornell.edu/~ross/publications/sleffects/sleffects-icfp18-tr.pdf) is not quite what you are after but certainly has interesting things to say about connecting monadic effects and comonadic coeffects. Hope those help. I would be interested in having a further conversation about this sometime if you want to grab a Skype call. Best, Dominic On Thu, 30 Apr 2020 at 07:24, Jonathan Aldrich wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear Types, > > I am curious about the relative expressiveness of effects and > coeffects. Has this been studied? > > The cleanest distinction I've seen is that effects capture the impact > a program has on its environment, i.e. what it produces. Coeffects > capture the requirements that a program puts on its environment: what > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > 2016 paper, "Combining Effects and Coeffects via Grading" (and > elsewhere). > > There is some useful intuition in this distinction, and it describes > the different structure of checking rules in effect and coeffect > systems. However, I don't find this distinction very helpful in > thinking about expressiveness. It seems like many examples can be > expressed in either an effect or a coeffect system. For example, an > exception is a classic example of an effect (e.g. in the paper > mentioned above, and many others). However, it seems to me that > exceptions can also be modeled as coeffects: code that might throw an > exception requires the caller to pass a handler for that exception to > it--or perhaps an abstract "permission" to throw that exception. So > in what sense are exceptions an effect, rather than a coeffect? > > Is this true of all the kinds of things that are typically expressed > with effects and coeffects--that they could just as easily be > expressed in the other style? If so, what are the benefits of one > style vs. the other? Or are there examples that can only be expressed > in one style--or for which expression in the other style is much more > awkward? > > Perhaps these questions have been written about, but I haven't been > able to find it. I would love to get some pointers. I am > particularly interested in a practical explanation of the differences > in expressiveness, or theoretical results that have a direct and > explicit relationship to practice--with practical examples in either > case. (co-)Monads and/or anything categorical are not a very helpful > starting point for me, but effect systems and/or linear types are > (I've done research on both). I am also more interested in the > descriptive view of effects/coeffects (in the sense of Filinski, > ICFP'11) than the prescriptive view. > > Thanks, > > Jonathan > From aleks.nanevski at imdea.org Thu Apr 30 13:21:49 2020 From: aleks.nanevski at imdea.org (Aleksandar Nanevski) Date: Thu, 30 Apr 2020 19:21:49 +0200 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: Hi Jonathan. In my PhD thesis (CMU 2004), I have considered the distinction between different kinds of effects using types from (contextual) modal logic S4. In S4 one has two different type constructors, Box (universal quantifier over possible worlds) and Diamond (existential quantifier). Computations with a Box type are ones that depend on the environment but don't change it. Box is a comonad. Computations with a Diamond type are ones that change it. Diamond is a monad. https://software.imdea.org/~aleks/thesis/CMU-CS-04-151.pdf The results in the thesis were based on the following papers and TR's. For example, this TR discusses that exceptions (and other control-flow effects such as delimited continuations) should be viewed as Box-ed (i.e., comonadic) computations https://software.imdea.org/~aleks/papers/effects/CMU-CS-03-149.pdf This paper discusses that state can be viewed as a Diamond (i.e., monadic) computation, but if one views it as a Boxed (comonadic) computation, then one still gets something familiar, namely, dynamic binding. The two kinds can also nicely interact in one and the same calculus. https://software.imdea.org/~aleks/papers/effects/possibility.pdf Another way to view these two different kinds of computation as follows. Once you put a contextual information onto the boxed type, that information serves as a *precondition* for your program. Once you put contextual information onto the diamond type, it serves as a *postcondition*. Having both together in a same system gives you Hoare Type Theory (in fact, a type-based version of separation logic), which is something that I worked on as a postdoc: https://software.imdea.org/~aleks/papers/hoarelogic/htt.pdf https://software.imdea.org/~aleks/papers/hoarelogic/icfp06.pdf I hope you find these pointers useful. Best regards, Aleks On 30/04/2020 01:48, Jonathan Aldrich wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Types, > > I am curious about the relative expressiveness of effects and > coeffects. Has this been studied? > > The cleanest distinction I've seen is that effects capture the impact > a program has on its environment, i.e. what it produces. Coeffects > capture the requirements that a program puts on its environment: what > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > 2016 paper, "Combining Effects and Coeffects via Grading" (and > elsewhere). > > There is some useful intuition in this distinction, and it describes > the different structure of checking rules in effect and coeffect > systems. However, I don't find this distinction very helpful in > thinking about expressiveness. It seems like many examples can be > expressed in either an effect or a coeffect system. For example, an > exception is a classic example of an effect (e.g. in the paper > mentioned above, and many others). However, it seems to me that > exceptions can also be modeled as coeffects: code that might throw an > exception requires the caller to pass a handler for that exception to > it--or perhaps an abstract "permission" to throw that exception. So > in what sense are exceptions an effect, rather than a coeffect? > > Is this true of all the kinds of things that are typically expressed > with effects and coeffects--that they could just as easily be > expressed in the other style? If so, what are the benefits of one > style vs. the other? Or are there examples that can only be expressed > in one style--or for which expression in the other style is much more > awkward? > > Perhaps these questions have been written about, but I haven't been > able to find it. I would love to get some pointers. I am > particularly interested in a practical explanation of the differences > in expressiveness, or theoretical results that have a direct and > explicit relationship to practice--with practical examples in either > case. (co-)Monads and/or anything categorical are not a very helpful > starting point for me, but effect systems and/or linear types are > (I've done research on both). I am also more interested in the > descriptive view of effects/coeffects (in the sense of Filinski, > ICFP'11) than the prescriptive view. > > Thanks, > > Jonathan From vikraman.choudhury at gmail.com Thu Apr 30 14:24:40 2020 From: vikraman.choudhury at gmail.com (Vikraman Choudhury) Date: Thu, 30 Apr 2020 19:24:40 +0100 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: <87pnbode6v.fsf@gmail.com> Hi Jonathan, I'd like to point out a recent paper[0] by Neel Krishnaswami and I, "Recovering Purity with Comonads and Capabilities" which provides a different take on this topic. The ability to perform an effect can be encoded by the use of permission/capability variables. A comonad/coeffect modality can be used to control access to capability variables. Using this idea, we can encode capability-safe and pure functions in an unsafe/impure calculus -- a pure function is a capability-safe function with no capabilities! We show that it is possible to use this comonad to embed the pure CBV lambda calculus into an impure calculus, while preserving the full beta-eta equational theory. This is not unlike the embedding of linear logics into intuitionistic logic. [0] https://arxiv.org/abs/1907.07283 Jonathan Aldrich writes: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Types, > > I am curious about the relative expressiveness of effects and > coeffects. Has this been studied? > > The cleanest distinction I've seen is that effects capture the impact > a program has on its environment, i.e. what it produces. Coeffects > capture the requirements that a program puts on its environment: what > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > 2016 paper, "Combining Effects and Coeffects via Grading" (and > elsewhere). > > There is some useful intuition in this distinction, and it describes > the different structure of checking rules in effect and coeffect > systems. However, I don't find this distinction very helpful in > thinking about expressiveness. It seems like many examples can be > expressed in either an effect or a coeffect system. For example, an > exception is a classic example of an effect (e.g. in the paper > mentioned above, and many others). However, it seems to me that > exceptions can also be modeled as coeffects: code that might throw an > exception requires the caller to pass a handler for that exception to > it--or perhaps an abstract "permission" to throw that exception. So > in what sense are exceptions an effect, rather than a coeffect? > > Is this true of all the kinds of things that are typically expressed > with effects and coeffects--that they could just as easily be > expressed in the other style? If so, what are the benefits of one > style vs. the other? Or are there examples that can only be expressed > in one style--or for which expression in the other style is much more > awkward? > > Perhaps these questions have been written about, but I haven't been > able to find it. I would love to get some pointers. I am > particularly interested in a practical explanation of the differences > in expressiveness, or theoretical results that have a direct and > explicit relationship to practice--with practical examples in either > case. (co-)Monads and/or anything categorical are not a very helpful > starting point for me, but effect systems and/or linear types are > (I've done research on both). I am also more interested in the > descriptive view of effects/coeffects (in the sense of Filinski, > ICFP'11) than the prescriptive view. > > Thanks, > > Jonathan -- Vikraman From daekharel at gmail.com Thu Apr 30 15:45:17 2020 From: daekharel at gmail.com (Michael Arntzenius) Date: Thu, 30 Apr 2020 21:45:17 +0200 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: Dear Jonathan, > Or are there examples that can only be expressed in one style--or for which expression in the other style is much more awkward? I was about to conjecture that I had an example of something that could only be expressed as a coeffect, but reading Dominic's reply I've realized I was wrong. I still offer it, though, as a concrete example of something that I think would be more awkward to express as an effect: non-monotonicity in an otherwise monotone world. I can only hope this isn't too obscure to be interesting :). The gist here is that, if we assign each of our types a partial order, and track for each function whether it is monotone with respect to this order or not in its type, then we can determine for each variable used in an expression whether it is used monotonically or not. This lets us track monotonicity in our type system. This involves associating information with the context ? namely how each variable is used ? so it is in essence a coeffect system. For full details you can read the ICFP 2016 paper "Datafun: A Functional Datalog". (This doesn't mention coeffects explicitly; I didn't notice the connection at the time.) I believe this is also similar to how variance/positivity checking in proof assistants works. It is also possible to express this as an effect, following the recipe laid out in ?Should I use a Monad or Comonad??. To do this, interpret our types as pre-ordered instead (i.e. drop the requirement that a type's order is antisymmetric). Then a non-monotone map A -> B can be represented by a monotone map A -> indisc B, where "indisc" is the "indiscreteness monad", which makes (x <= y : indisc B) always true! So this makes non-monotonicity into an effect; in essence, purity = monotonicity, impurity = non-monotonicity. But this approach makes working with functions that are monotone in some but not all arguments awkward. For example, a function f : A -> B -> C that is monotone in B but not A must be typed A -> indisc (B -> C). This is not equivalent to A -> B -> indisc C, which does not guarantee monotonicity in B. From an effect-based POV, the former requires the "effect" to occur before the function has even seen the value of type B. With most effects we rarely care about details like this, but in Datafun we care a lot. So for our use case, I think a context-oriented coeffect view is more convenient for the programmer. In wonder if this generalizes? Perhaps, if you find yourself working with values of type T(A -> B) often, where T is a monad, it might be more convenient to work with an equivalent coeffect system instead, should one exist. Cheers, Michael Arntzenius On Thu, Apr 30, 2020 at 18:14 Dominic Orchard wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear Jonathan, > > This is a great question and something that I think still lacks a thorough > comprehensive treatment in the literature, but certainly there are hints > about this in various places (and I have been slowly feeling my way towards > different versions of this question, including some work in progress right > now!) > > This is certainly something that we are trying to explore in the Granule > project (https://granule-project.github.io/, and at ICFP 2019) where we > have a fully functioning modern functional programming language which can > work with various different effects and coeffects at the same time. We are > brewing a few things comparing certain analyses performed as effects vs > coeffects (because, as you say, it seems that some things can be done both > ways, so why choose one over the other?). It's probably worth pointing out > that, whilst there is quite a bit of recent work with coeffects in a linear > type theory (like Granule and the ICFP 2016 paper you mention), coeffects > need not be tided to a linear type theory (indeed, the first two papers > that Tomas, Alan, and I wrote on this were in a Cartesian setting (ICALP > 2013 and ICFP 2014)). > > My general, informal view is that (broadly) monadic effects capture a kind > of 'forwards' analysis, with information flowing from inputs to outputs, > whereas comonadic coeffects capture an 'backwards'-style analysis, flowing > back through the inputs. How you structure the primitives for working with > either will affect how easy it is to capture certain properties (and the > CBN/CBV story then becomes important as well, see the Cicek et al. paper > mentioned below). > > Here are a few references that spring to mind that go a little bit towards > your question: > > * Tomas Petricek and I reflected on coeffect vs effect expressivity a bit > in Section 8 of this paper ( > https://www.cs.kent.ac.uk/people/staff/dao7/publ/haskell14-effects.pdf) > where we give a brief comparison between expressing implicit parameters as > effects vs coeffects (reader graded monad vs product graded comonad) and > what the difference is (we point out that Haskell's implicit parameters > feature are a coeffect rather than an effect, perhaps the most "mainstream" > example of explicit coeffects in the wild!). > > * This short paper by Ezgi Cicek, Marco Gaboardi, and Deepak Garg looks at > this question in the context of cost analysis ( > https://lipn.univ-paris13.fr/DICE2016/Abstracts/paper_10.pdf). > > * I know that you said a (co)monad treatment is not helpful to you, but in > case it helps others, I have this unpublished article from the end of my > PhD "Should I use a Monad or Comonad?" which at least explains some overlap > ( > > https://www.cs.kent.ac.uk/people/staff/dao7/drafts/monad-or-comonad-orchard11-draft.pdf > ) > and tries to reconcile some tensions in the literature. The paper is meant > to be a fairly self-contained. > > * The ICFP 2018 by Andrew Hirsch and Ross Tate ( > > http://www.cs.cornell.edu/~ross/publications/sleffects/sleffects-icfp18-tr.pdf > ) > is not quite what you are after but certainly has interesting things to say > about connecting monadic effects and comonadic coeffects. > > Hope those help. I would be interested in having a further conversation > about this sometime if you want to grab a Skype call. > > Best, > Dominic > > On Thu, 30 Apr 2020 at 07:24, Jonathan Aldrich < > jonathan.aldrich at cs.cmu.edu> > wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > Dear Types, > > > > I am curious about the relative expressiveness of effects and > > coeffects. Has this been studied? > > > > The cleanest distinction I've seen is that effects capture the impact > > a program has on its environment, i.e. what it produces. Coeffects > > capture the requirements that a program puts on its environment: what > > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > > 2016 paper, "Combining Effects and Coeffects via Grading" (and > > elsewhere). > > > > There is some useful intuition in this distinction, and it describes > > the different structure of checking rules in effect and coeffect > > systems. However, I don't find this distinction very helpful in > > thinking about expressiveness. It seems like many examples can be > > expressed in either an effect or a coeffect system. For example, an > > exception is a classic example of an effect (e.g. in the paper > > mentioned above, and many others). However, it seems to me that > > exceptions can also be modeled as coeffects: code that might throw an > > exception requires the caller to pass a handler for that exception to > > it--or perhaps an abstract "permission" to throw that exception. So > > in what sense are exceptions an effect, rather than a coeffect? > > > > Is this true of all the kinds of things that are typically expressed > > with effects and coeffects--that they could just as easily be > > expressed in the other style? If so, what are the benefits of one > > style vs. the other? Or are there examples that can only be expressed > > in one style--or for which expression in the other style is much more > > awkward? > > > > Perhaps these questions have been written about, but I haven't been > > able to find it. I would love to get some pointers. I am > > particularly interested in a practical explanation of the differences > > in expressiveness, or theoretical results that have a direct and > > explicit relationship to practice--with practical examples in either > > case. (co-)Monads and/or anything categorical are not a very helpful > > starting point for me, but effect systems and/or linear types are > > (I've done research on both). I am also more interested in the > > descriptive view of effects/coeffects (in the sense of Filinski, > > ICFP'11) than the prescriptive view. > > > > Thanks, > > > > Jonathan > > > From neelakantan.krishnaswami at gmail.com Thu Apr 30 17:59:03 2020 From: neelakantan.krishnaswami at gmail.com (Neel Krishnaswami) Date: Thu, 30 Apr 2020 22:59:03 +0100 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: <87pnbode6v.fsf@gmail.com> References: <87pnbode6v.fsf@gmail.com> Message-ID: Hello, I sent most of this message to Jonathan privately, but after Vikraman's comment I realized that most of it is surely appropriate here, as well. First, I want to distinguish between two sorts of effect systems, descriptive effect systems and synthetic effect systems. Descriptive effect systems start with the idea that the language is describing an underlying semantics or abstract machine, and the effect annotations describe which subset of the effects are allowed to happen. On the other hand, synthetic effect systems are about building custom effects/monads for a program -- effect handlers are an example of this kind. I am going to restrict my remarks to the first kind of effect system, focusing particularly to capabilities. In this setting, a machine may have many possible effects -- for example, we can imagine count writing each file a program to write to as a different effect. An effect system in the classic sense is basically a family of monads, one per subset of allowed effects. That is, we can imagine a family of monads ??? T(C, X) where C is the set of channels that the computation is permitted to write on, and X is the return type. If we take the indexing by allowed write set seriously, we get a *graded monad*. That is, the monadic operations get the refined types: ??? return : X ? T(?, X) ??? join?? : T(C, T(C', X)) ? T(C ? C', X) The first says that `return` produces a computation that will not write to any channels, and the fact that I wrote an isomorphism means you can escape computations which don't perform any effects. The type of `join` says that if you have a computation that could write to `C`, and returns a computation that could write to `C'`, then when you sequence the computations to get the `X`, you could have writes on `C ? C'`. To get a precise type for the `write` operation, you need a little bit of indexed typing (Dominic Orchard & co's Granule language could certainly express this, for instance): ??? write : ?c. Channel(c) ? String ? T({c}, unit) This says that the `print` function takes the channel `c` and a string, and performs a computation that may write on the singleton set of channels `{c}`. Now that we have a type of channel values, it becomes natural to think of values as *owning* capabilities. If you have a type `Channel(c)` and `Channel(c')` then we expect the linear pair type `Channel(c) ? Channel(c')` to represent the ownership of two *distinct* channels. This is very much like the `P ? Q` separating conjunction in separation logic. Similarly, there should be a linear function space `X ? Y` representing a function receiving an argument with capabilities disjoint from the function, much like the magic wand of separation logic. How should tensor products and computations interact? Well, we know that the monad and tensor should satisfy the property of having a *strength*: ??? strength : T(C, X) ? Y ? T(C, X ? Y) Surprisingly, having a linear strength corresponds to the *frame property* of separation logic. Intuitively, if you think of ??? f : P ? T(C, Q) as representing a computation that has a precondition `P` and postcondition `Q`, then having a strength is what you need to derive ??? g : P ? R ? T(C, Q ? R) You can see it with the following linear term: ??? frame : (P ? T(C, Q)) ? (P ? R) ? T(C, Q ? R) ??? frame f (p, r) = strength(f p, r) I documented this relationship in my POPL 2015 paper, *Integrating Linear and Dependent Types*, but I think it was folklore before that: Bob Atkey mentioned to me that he knew it as well.? In addition, I should note having a strength with the tensor is what is needed to make the linear version of do-notation work out the way that you expect. Now, once you have a linear type discipline for capabilities, it is natural to ask what the exponential !A should mean. Brunel, Gaboardi, Mazza and Zdancewic suggested in their 2014 paper *A Core Quantitative Coeffect Calculus* that coeffects can be understood as *graded exponentials* -- just as an effect system can be seen as graded monad, graded comonads can model coeffects. This is a bit abstract, so it is worth asking what kinds of graded comonads arise in the specific context of capabilities. One interested one is the *safety comonad*. This is, we introduce a family of graded comonads: ??? Safe(C, X) which represent expressions of type `X` which **do not** own any capabilities in the `C`. Notice the duality here -- an effect type `T(C, X)` says that the computation *may use some* of the capabilities in `C`, and the safety comonad `Safe(C, X)` says that the expression *must not own any* capabilities in `C`. There satisfy the comonad laws for the evident operations: ??? extract : Safe(C, X) ? X ??? duplicate : Safe(C, X) ? Safe(C, Safe(C, X)) which means that if you have an `X` which owns nothing in `C`, it is still an `X`, and if you have a value which does not own `C`, and that a value which does not own `C` does not own `C`. The `Safe` comonad also satisfies the properties you might expect given the intuition about denial of ownership: ??? iso? : Safe(?, X) ? X ??? dist : Safe(C ? C', X) ? Safe(C, Safe(C', X)) I should note that this is not exactly the definition of graded comonad given in the literature, but IMO that's not enormously surprising -- the exponential in linear logic is not canonical (you can have multiple non-isomorphic exponentials for the same linear language), and so you have to let the application guide you. Now, two surprising things happen. First, in this setting, there is a natural notion of nonlinear arrow which **is not** the linear decomposition `!A ? B`. Instead, there is a free-standing function space `A ? B` which represents a function which can *share* capabilities with its argument, just like the ordinary implication in separation logic permits sharing resources between hypothesis and consequent. Indeed, this notion of function type corresponds to *capability-safe functions* -- these are functions which can only access capabilities either through their argument or their free variables. That is, they are functions which *lack ambient authority*. So we have two notions of function: those which can transfer ownership and have a linear type discipline, and those which are capability-safe and have a nonlinear type discipline. (Together, they form a categorical model of bunched implications.) Second, the comonad and the monad have an interaction: ??? cancel : Safe(C, T(C', X)) ? Safe(C, T(C' - C, X)) The type of the cancellation operation `cancel` says that if you have a computation which does not own any capabilities in `C`, then the computation cannot perform any writes to channels in `C`. In particular, if `C` is bigger than `C'`, then we get ??? cancel : Safe(C, T(C', X)) ? Safe(C, T(C' - C, X)) ?????????? = Safe(C, T(C', X)) ? Safe(C, T(?, X))????? // because C ??? ??? ?? ? Safe(C, T(C', X)) ? Safe(C, X) which says that you can "escape the monad" if you know that you don't own any of the capabilities the computation may use.? In my view, this cancellation property is the essence of how effects and coeffects interact in the setting of capabilities. Interestingly, this cancellation law does not appear to be an instance of the kinds of distributive law that Marco Gaboardi, Shin-ya Katsumata, Dominic Orchard, Flavien Breuvart, and Tarmo Uustalu studied in their ICFP 2016 paper, *Combining Effects and Coeffects via Grading*. (As Dominic and Flavien remarked earlier in the thread, there is a lot of unexplored territory in this space.) Specifically in the context of capabilities, neither effects nor coeffects supplant the other, and neither do ownership transfer nor capability-safety supplant the other! But put together they give you a very fine-grained ability to control and reason about how capabilities are transferred and used. Best, Neel On 30/04/2020 19:24, Vikraman Choudhury wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > Hi Jonathan, > > I'd like to point out a recent paper[0] by Neel Krishnaswami and I, > "Recovering Purity with Comonads and Capabilities" which provides a > different take on this topic. > > The ability to perform an effect can be encoded by the use of > permission/capability variables. A comonad/coeffect modality can be used > to control access to capability variables. Using this idea, we can > encode capability-safe and pure functions in an unsafe/impure calculus > -- a pure function is a capability-safe function with no capabilities! > > We show that it is possible to use this comonad to embed the pure CBV > lambda calculus into an impure calculus, while preserving the full > beta-eta equational theory. This is not unlike the embedding of linear > logics into intuitionistic logic. > > [0] https://arxiv.org/abs/1907.07283 > > Jonathan Aldrich writes: > >> [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >> >> Dear Types, >> >> I am curious about the relative expressiveness of effects and >> coeffects. Has this been studied? >> >> The cleanest distinction I've seen is that effects capture the impact >> a program has on its environment, i.e. what it produces. Coeffects >> capture the requirements that a program puts on its environment: what >> it consumes. This is discussed, for example, in Gaboardi et al's ICFP >> 2016 paper, "Combining Effects and Coeffects via Grading" (and >> elsewhere). >> >> There is some useful intuition in this distinction, and it describes >> the different structure of checking rules in effect and coeffect >> systems. However, I don't find this distinction very helpful in >> thinking about expressiveness. It seems like many examples can be >> expressed in either an effect or a coeffect system. For example, an >> exception is a classic example of an effect (e.g. in the paper >> mentioned above, and many others). However, it seems to me that >> exceptions can also be modeled as coeffects: code that might throw an >> exception requires the caller to pass a handler for that exception to >> it--or perhaps an abstract "permission" to throw that exception. So >> in what sense are exceptions an effect, rather than a coeffect? >> >> Is this true of all the kinds of things that are typically expressed >> with effects and coeffects--that they could just as easily be >> expressed in the other style? If so, what are the benefits of one >> style vs. the other? Or are there examples that can only be expressed >> in one style--or for which expression in the other style is much more >> awkward? >> >> Perhaps these questions have been written about, but I haven't been >> able to find it. I would love to get some pointers. I am >> particularly interested in a practical explanation of the differences >> in expressiveness, or theoretical results that have a direct and >> explicit relationship to practice--with practical examples in either >> case. (co-)Monads and/or anything categorical are not a very helpful >> starting point for me, but effect systems and/or linear types are >> (I've done research on both). I am also more interested in the >> descriptive view of effects/coeffects (in the sense of Filinski, >> ICFP'11) than the prescriptive view. >> >> Thanks, >> >> Jonathan > > -- > Vikraman From marco.servetto at gmail.com Fri May 1 00:57:33 2020 From: marco.servetto at gmail.com (Marco Servetto) Date: Fri, 1 May 2020 16:57:33 +1200 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: Message-ID: Hi, Citing Michael "This involves associating information with the context ? namely how each variable is used ? so it is in essence a coeffect system." This indeed seams to confirm my general understanding: effects are computed\inferred coeffects are checked So, is every kind of "type environment" is a form of a coeffect? in that case, here we have the answer to the question of Jonathan: Lets consider one of the simplest and most famous type systems: the simply typed lambda calculus. We can encode the type of variables in an environment Gamma (x1:T1..xn:Tn) and we can get the simple and elegant type system we all known. OR we can make a complex system of constraints and we can have a Judgment of the form |-e: and we would type the lambda abstraction |-e : ---------------------------------------------------- |- \x:T1.e : < T1->T : Gamma[x=T1] > Then, depending on how we define types T, Gamma and Gamma[x=T] we would either have a mathematically equivalent version of the conventional system but "wrote funny" or a highly generic type system where the type needed for local variables is expressed as constraints and Gamma[x=T1] is adding more constraints (possibly triggering a simplification-minimization of the current constraint set) In general, you get a form of type inference from effect systems: Imagine if such simply typed lambda calculus was to ALSO support an "untyped" lambda \x.e; with an effect type system, such lambda term would be able to work as a "generic method". On the other side, effects tend to be much harder to define\compute then simply checking information readily available in an environment. Going back to the original case of the exceptions: if all the methods define what exceptions they are allowed to leak, then effects are redundant effort. On the other side, if we wanted to infer what exceptions a method can throw, in the same way the simply typed lambda calculus infers the return type of lambda terms, then effects would be perfect for the task. Marco! On Fri, 1 May 2020 at 09:04, Michael Arntzenius wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Jonathan, > > > Or are there examples that can only be expressed > in one style--or for which expression in the other style is much more > awkward? > > I was about to conjecture that I had an example of something that could > only be expressed as a coeffect, but reading Dominic's reply I've realized > I was wrong. I still offer it, though, as a concrete example of something > that I think would be more awkward to express as an effect: > non-monotonicity in an otherwise monotone world. I can only hope this isn't > too obscure to be interesting :). > > The gist here is that, if we assign each of our types a partial order, and > track for each function whether it is monotone with respect to this order > or not in its type, then we can determine for each variable used in an > expression whether it is used monotonically or not. This lets us track > monotonicity in our type system. This involves associating information with > the context ? namely how each variable is used ? so it is in essence a > coeffect system. > > For full details you can read the ICFP 2016 paper "Datafun: A Functional > Datalog". (This doesn't mention coeffects explicitly; I didn't notice the > connection at the time.) I believe this is also similar to how > variance/positivity checking in proof assistants works. > > It is also possible to express this as an effect, following the recipe laid > out in ?Should I use a Monad or Comonad??. To do this, interpret our types > as pre-ordered instead (i.e. drop the requirement that a type's order is > antisymmetric). Then a non-monotone map A -> B can be represented by a > monotone map A -> indisc B, where "indisc" is the "indiscreteness monad", > which makes (x <= y : indisc B) always true! So this makes non-monotonicity > into an effect; in essence, purity = monotonicity, impurity = > non-monotonicity. > > But this approach makes working with functions that are monotone in some > but not all arguments awkward. For example, a function f : A -> B -> C that > is monotone in B but not A must be typed A -> indisc (B -> C). This is not > equivalent to A -> B -> indisc C, which does not guarantee monotonicity in > B. From an effect-based POV, the former requires the "effect" to occur > before the function has even seen the value of type B. With most effects we > rarely care about details like this, but in Datafun we care a lot. So for > our use case, I think a context-oriented coeffect view is more convenient > for the programmer. > > In wonder if this generalizes? Perhaps, if you find yourself working with > values of type T(A -> B) often, where T is a monad, it might be more > convenient to work with an equivalent coeffect system instead, should one > exist. > > Cheers, > Michael Arntzenius > > On Thu, Apr 30, 2020 at 18:14 Dominic Orchard wrote: > > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > Dear Jonathan, > > > > This is a great question and something that I think still lacks a thorough > > comprehensive treatment in the literature, but certainly there are hints > > about this in various places (and I have been slowly feeling my way towards > > different versions of this question, including some work in progress right > > now!) > > > > This is certainly something that we are trying to explore in the Granule > > project (https://granule-project.github.io/, and at ICFP 2019) where we > > have a fully functioning modern functional programming language which can > > work with various different effects and coeffects at the same time. We are > > brewing a few things comparing certain analyses performed as effects vs > > coeffects (because, as you say, it seems that some things can be done both > > ways, so why choose one over the other?). It's probably worth pointing out > > that, whilst there is quite a bit of recent work with coeffects in a linear > > type theory (like Granule and the ICFP 2016 paper you mention), coeffects > > need not be tided to a linear type theory (indeed, the first two papers > > that Tomas, Alan, and I wrote on this were in a Cartesian setting (ICALP > > 2013 and ICFP 2014)). > > > > My general, informal view is that (broadly) monadic effects capture a kind > > of 'forwards' analysis, with information flowing from inputs to outputs, > > whereas comonadic coeffects capture an 'backwards'-style analysis, flowing > > back through the inputs. How you structure the primitives for working with > > either will affect how easy it is to capture certain properties (and the > > CBN/CBV story then becomes important as well, see the Cicek et al. paper > > mentioned below). > > > > Here are a few references that spring to mind that go a little bit towards > > your question: > > > > * Tomas Petricek and I reflected on coeffect vs effect expressivity a bit > > in Section 8 of this paper ( > > https://www.cs.kent.ac.uk/people/staff/dao7/publ/haskell14-effects.pdf) > > where we give a brief comparison between expressing implicit parameters as > > effects vs coeffects (reader graded monad vs product graded comonad) and > > what the difference is (we point out that Haskell's implicit parameters > > feature are a coeffect rather than an effect, perhaps the most "mainstream" > > example of explicit coeffects in the wild!). > > > > * This short paper by Ezgi Cicek, Marco Gaboardi, and Deepak Garg looks at > > this question in the context of cost analysis ( > > https://lipn.univ-paris13.fr/DICE2016/Abstracts/paper_10.pdf). > > > > * I know that you said a (co)monad treatment is not helpful to you, but in > > case it helps others, I have this unpublished article from the end of my > > PhD "Should I use a Monad or Comonad?" which at least explains some overlap > > ( > > > > https://www.cs.kent.ac.uk/people/staff/dao7/drafts/monad-or-comonad-orchard11-draft.pdf > > ) > > and tries to reconcile some tensions in the literature. The paper is meant > > to be a fairly self-contained. > > > > * The ICFP 2018 by Andrew Hirsch and Ross Tate ( > > > > http://www.cs.cornell.edu/~ross/publications/sleffects/sleffects-icfp18-tr.pdf > > ) > > is not quite what you are after but certainly has interesting things to say > > about connecting monadic effects and comonadic coeffects. > > > > Hope those help. I would be interested in having a further conversation > > about this sometime if you want to grab a Skype call. > > > > Best, > > Dominic > > > > On Thu, 30 Apr 2020 at 07:24, Jonathan Aldrich < > > jonathan.aldrich at cs.cmu.edu> > > wrote: > > > > > [ The Types Forum, > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > ] > > > > > > Dear Types, > > > > > > I am curious about the relative expressiveness of effects and > > > coeffects. Has this been studied? > > > > > > The cleanest distinction I've seen is that effects capture the impact > > > a program has on its environment, i.e. what it produces. Coeffects > > > capture the requirements that a program puts on its environment: what > > > it consumes. This is discussed, for example, in Gaboardi et al's ICFP > > > 2016 paper, "Combining Effects and Coeffects via Grading" (and > > > elsewhere). > > > > > > There is some useful intuition in this distinction, and it describes > > > the different structure of checking rules in effect and coeffect > > > systems. However, I don't find this distinction very helpful in > > > thinking about expressiveness. It seems like many examples can be > > > expressed in either an effect or a coeffect system. For example, an > > > exception is a classic example of an effect (e.g. in the paper > > > mentioned above, and many others). However, it seems to me that > > > exceptions can also be modeled as coeffects: code that might throw an > > > exception requires the caller to pass a handler for that exception to > > > it--or perhaps an abstract "permission" to throw that exception. So > > > in what sense are exceptions an effect, rather than a coeffect? > > > > > > Is this true of all the kinds of things that are typically expressed > > > with effects and coeffects--that they could just as easily be > > > expressed in the other style? If so, what are the benefits of one > > > style vs. the other? Or are there examples that can only be expressed > > > in one style--or for which expression in the other style is much more > > > awkward? > > > > > > Perhaps these questions have been written about, but I haven't been > > > able to find it. I would love to get some pointers. I am > > > particularly interested in a practical explanation of the differences > > > in expressiveness, or theoretical results that have a direct and > > > explicit relationship to practice--with practical examples in either > > > case. (co-)Monads and/or anything categorical are not a very helpful > > > starting point for me, but effect systems and/or linear types are > > > (I've done research on both). I am also more interested in the > > > descriptive view of effects/coeffects (in the sense of Filinski, > > > ICFP'11) than the prescriptive view. > > > > > > Thanks, > > > > > > Jonathan > > > > > From jung at mpi-sws.org Fri May 1 06:57:16 2020 From: jung at mpi-sws.org (Ralf Jung) Date: Fri, 1 May 2020 12:57:16 +0200 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: References: <87pnbode6v.fsf@gmail.com> Message-ID: <50b30f15-ef3f-fdcf-2491-836b8c76b8c4@mpi-sws.org> Hallo Neel, (This is moving away from effects and coeeffects a bit, I hope that is okay.) > How should tensor products and computations interact? Well, we > know that the monad and tensor should satisfy the property of > having a *strength*: > > ??? strength : T(C, X) ? Y ? T(C, X ? Y) > > Surprisingly, having a linear strength corresponds to the *frame > property* of separation logic. Intuitively, if you think of > > ??? f : P ? T(C, Q) > > as representing a computation that has a precondition `P` and > postcondition `Q`, then having a strength is what you need to > derive > > ??? g : P ? R ? T(C, Q ? R) > > You can see it with the following linear term: > > ??? frame : (P ? T(C, Q)) ? (P ? R) ? T(C, Q ? R) > ??? frame f (p, r) = strength(f p, r) > > I documented this relationship in my POPL 2015 paper, *Integrating > Linear and Dependent Types*, but I think it was folklore before that: > Bob Atkey mentioned to me that he knew it as well.? In addition, I > should note having a strength with the tensor is what is needed to > make the linear version of do-notation work out the way that you > expect. Indeed we (later, independently) made the same observation for the laws of our "update modality" in Iris [0], where I wrote a rule like your "strength" above but called it "framing" (not knowing monad terminology), and then someone pointed out the connection to me. [0]: https://www.mpi-sws.org/~dreyer/papers/iris-ground-up/paper.pdf > Now, two surprising things happen. > > First, in this setting, there is a natural notion of nonlinear arrow > which **is not** the linear decomposition `!A ? B`. Instead, there is > a free-standing function space `A ? B` which represents a function > which can *share* capabilities with its argument, just like the > ordinary implication in separation logic permits sharing resources > between hypothesis and consequent. In Iris, being a separation logic, we also have both magic wand and implication. I have long wondered about the connection to linear logic and was quite surprised when I realized that the "implication" in linear logic (`!A ? B`) does not correspond to our implication in Iris. Ever since then I wonder why implication in linear logic looks the way it does. > Indeed, this notion of function type corresponds to *capability-safe > functions* -- these are functions which can only access capabilities > either through their argument or their free variables. That is, they > are functions which *lack ambient authority*. This is very interesting. While we *have* both magic wand and implication, we don't actually use implication in Iris. (We do use the fact that existentials commute with conjunction, which implies the existence of implication, but that's it.) We only use magic wand. Both are equivalent in some situations (but this, I think, relies on the fact that Iris is linear and not affine), which is why you might see implications in our papers, but really we only do that to hopefully not confuse the reader with too many magic wands. This is in contrast to the conjunctions, where we do use both of them in meaningful ways. So in your context of doing this with a linear programming language, do you have an example where implication is actually *useful*? As you mention above, these functions can access their arguments *or* their environment, but not both, which seems rather restrictive. (I do not understand what you mean by "lacking ambient authority".) Kind regards, Ralf From p.ohearn at ucl.ac.uk Fri May 1 09:54:24 2020 From: p.ohearn at ucl.ac.uk (O'Hearn, Peter) Date: Fri, 1 May 2020 13:54:24 +0000 Subject: [TYPES] effect vs. coeffect expressiveness In-Reply-To: <50b30f15-ef3f-fdcf-2491-836b8c76b8c4@mpi-sws.org> References: <87pnbode6v.fsf@gmail.com> <50b30f15-ef3f-fdcf-2491-836b8c76b8c4@mpi-sws.org> Message-ID: <0D9B408E-EC00-4B6B-8E30-E3246968CECC@ucl.ac.uk> On 1 May 2020, at 11:57, Ralf Jung > wrote: In Iris, being a separation logic, we also have both magic wand and implication. I have long wondered about the connection to linear logic and was quite surprised when I realized that the "implication" in linear logic (`!A ? B`) does not correspond to our implication in Iris. Ever since then I wonder why implication in linear logic looks the way it does. Hi Ralf, the difference between BI and LL was explained in the very first paper on BI by Pym and I: see the last section of http://www0.cs.ucl.ac.uk/staff/p.ohearn/papers/BI.pdf . SL inherits the differences. The most important point (IMHO) is that if you take a semantic perspective, then the difference jumps out so should not be surprising. And I think looking at the models explains ?why? each of the implications is like they are: they flow inevitably from natural models. In essence, ? BI corresponds to a single category possessing both a monoidal closed (so, -*) and a cartesian closed (so, ->) structure **in the same category**. Naturally occurring model: presheaves over a symmetric monoidal category of worlds (like in separation logic, or rather Day?s pro-monoidal structure). ? (intuitionistic) LL is semantically two categories, one monoidal closed and the other cartesian closed, with an adjunction between them (which gives you !). So yes two closed structure (two implications), but **not for the same category**. Naturally occurring model: coherence spaces and linear maps as SMCC, coherence spaces and stable maps as CCC (I hope I remember correctly). You can also choose to induce the second category as the Klesli category of a comonad, but I find the ?two categories? model more natural. BI also has a natural Kripke truth-functional semantics which you are used to, but I have always drawn LL intuition not from a truth-functional semantics but from its lovely semantics of proofs (like coherence spaces). The LL monoidal closed category is usually not also cartesian closed. E.g., coherence spaces and linear maps case. There is no value judgement in these models: the two systems just fit their models very naturally, and the models occur in nature. If you don?t have one of these models in your hands, no problem, no need to apply a shoehorn. Anyhow, once on sees the semantics of BI and LL in these terms, I hope it is not surprising that LL does not ?have? -> (if you are sitting in the monoidal category, like you are the way LL is usually formulated). You can replace ?monoidal closed category? by ?residuated monoid? and CCC by ?heyting algebra? for a non-categorical explanation of the above. My paper ?On bunched Typing? (http://www.cs.ucl.ac.uk/staff/p.ohearn/papers/BunchedTyping.pdf) contrasts they systems? views of shared versus consumable resources (in old OS terminology); i.e., using types instead of models. best, Peter From selinger at mathstat.dal.ca Fri May 1 23:53:11 2020 From: selinger at mathstat.dal.ca (selinger at mathstat.dal.ca) Date: Sat, 2 May 2020 00:53:11 -0300 (ADT) Subject: [TYPES] Towards Sustainable Open Access: thoughts on the ACM OPEN In-Reply-To: Message-ID: Dear type theorists, some months ago, we had a discussion about the ACM and open access. If you are interested in this issue and you are an ACM member, I'd like to draw your attention to the fact that the ACM currently has an election going on, for everything from president downwards. You will have received an email with instructions on how to vote electronically. It is very easy to look through the list of candiates (even using CTRL-F) and see who said something about open access in their statements. Definitely not all candidates have this on their radar. The ones that do can either be paraphrased as "we did a great job with open access" or "we need to push for better open access". -- Peter From rnrand at gmail.com Sun May 3 00:10:19 2020 From: rnrand at gmail.com (Robert Rand) Date: Sun, 3 May 2020 00:10:19 -0400 Subject: [TYPES] Towards Sustainable Open Access: thoughts on the ACM OPEN In-Reply-To: References: Message-ID: I'd add that I reached out to all of the candidates on behalf of the signatories to the OpenACM petition, to get details on their positions on open access. All of the responses are available in the Google doc below and I'd encourage you to read it before voting in the election (and please do vote!). You should have received information on voting in an email with the subject line "2020 ACM General Election". https://docs.google.com/document/d/1Oeafn8_N00fW1Exi1K-x7Jw6Pc0_ElZv7XZBF8soTBE/edit?usp=sharing Best, Robert On Sat, May 2, 2020 at 3:16 AM wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear type theorists, > > some months ago, we had a discussion about the ACM and open access. > > If you are interested in this issue and you are an ACM member, I'd > like to draw your attention to the fact that the ACM currently has an > election going on, for everything from president downwards. You will > have received an email with instructions on how to vote > electronically. It is very easy to look through the list of candiates > (even using CTRL-F) and see who said something about open access in > their statements. Definitely not all candidates have this on their > radar. The ones that do can either be paraphrased as "we did a great > job with open access" or "we need to push for better open access". > > -- Peter > From roberto at dicosmo.org Mon May 4 08:47:48 2020 From: roberto at dicosmo.org (Roberto Di Cosmo) Date: Mon, 4 May 2020 14:47:48 +0200 Subject: [TYPES] [Announce]: biblatex-software package available for citing software artifacts Message-ID: Dear all, software is a pillar of modern research, and yet, no support for citing software projects was available in any common bibliography style. Change is coming: this message is to announce the availability of the *biblatex-software* package, that adds support for properly managing software entries in bibliographies for BibLaTeX You can find it on CTAN at: https://www.ctan.org/tex-archive/macros/latex/contrib/biblatex-contrib/biblatex-software *biblatex-software* introduces *four specific bibliographic entries* for describing respectively *software, software versions, software modules and code fragments*, designed by a dedicated task force at Inria composed of Pierre Alliez, Roberto Di Cosmo, Benjamin Guedj, Alain Girault, Mohand-Said Hacid, Arnaud Legrand, Morane Gruenpeter, Xavier Leroy, Nicolas Rougier and Manuel Serrano. This is actually a bibliography *style* *extension*, which means that one can *add support for these four software entry types to any other BibLaTeX style* used in documents typeset in LaTeX, including in particular the biblatex style for ACM articles . It requires the Biber back end, though, so do not try it with pure BibTeX. Full documentation and examples are included in the package, see in particular - the style documentation http://mirrors.ctan.org/macros/latex/contrib/biblatex-contrib/biblatex-software/software-biblatex.pdf - an example document using this style http://mirrors.ctan.org/macros/latex/contrib/biblatex-contrib/biblatex-software/sample-use-sty.pdf - an example BibTeX file showcasing the four entries http://mirrors.ctan.org/macros/latex/contrib/biblatex-contrib/biblatex-software/biblio.bib Feel free to use this style, and forward this message to anybody interested. Contributions are welcome, in particular in the form of localization strings (English and French are already done) on https://gitlab.inria.fr/gt-sw-citation/bibtex-sw-entry Cheers -- Roberto ------------------------------------------------------------------ Computer Science Professor (on leave at INRIA from IRIF/Universit? de Paris) Director Software Heritage E-mail : roberto at dicosmo.org INRIA Web : http://www.dicosmo.org Bureau C328 Twitter : http://twitter.com/rdicosmo 2, Rue Simone Iff Tel : +33 1 80 49 44 42 CS 42112 75589 Paris Cedex 12 ------------------------------------------------------------------ GPG fingerprint 2931 20CE 3A5A 5390 98EC 8BFC FCCA C3BE 39CB 12D3 From alan.schmitt at polytechnique.org Fri Jun 19 10:06:44 2020 From: alan.schmitt at polytechnique.org (Alan Schmitt) Date: Fri, 19 Jun 2020 16:06:44 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? Message-ID: <87sger6tfv.fsf@m4x.org> Hello, I'm not sure this is the right place to ask this question, but I don't know of a forum where I could reach as many researchers who might have the same problem. Please tell me if there is a better place. I want to submit a paper to a conference that uses light double-blind reviewing. A crucial part of the work described by the paper is a piece of software, and I?feel that it would be difficult to assess the paper without being able to run the software, simply to check that it does what we claim (this is what a reviewer said in a previous attempt). So I want to anonymize the code and include it as an artefact. But that means I need to remove all the copyright info, which bothers me a little. Should I just not worry about it, or are there better ways to share artifacts anonymously? Thanks, Alan From mukeshtiwari.iiitm at gmail.com Sat Jun 20 09:36:43 2020 From: mukeshtiwari.iiitm at gmail.com (mukesh tiwari) Date: Sat, 20 Jun 2020 23:36:43 +1000 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87sger6tfv.fsf@m4x.org> References: <87sger6tfv.fsf@m4x.org> Message-ID: Hi Alan, You can use https://anonymous.4open.science/ for sharing your code anonymously. We (me and my coauthors) have used it in the past, but there were no copyright issues with us. Try uploading your source code, and see how it anonymises your code. Best, Mukesh On Sat, Jun 20, 2020 at 11:10 PM Alan Schmitt < alan.schmitt at polytechnique.org> wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Hello, > > I'm not sure this is the right place to ask this question, but I don't > know of a forum where I could reach as many researchers who might have > the same problem. Please tell me if there is a better place. > > I want to submit a paper to a conference that uses light double-blind > reviewing. A crucial part of the work described by the paper is a piece > of software, and I feel that it would be difficult to assess the paper > without being able to run the software, simply to check that it does > what we claim (this is what a reviewer said in a previous attempt). So I > want to anonymize the code and include it as an artefact. But that means > I need to remove all the copyright info, which bothers me a little. > Should I just not worry about it, or are there better ways to share > artifacts anonymously? > > Thanks, > > Alan > From eeide at cs.utah.edu Sat Jun 20 10:20:05 2020 From: eeide at cs.utah.edu (Eric Eide) Date: Sat, 20 Jun 2020 08:20:05 -0600 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87sger6tfv.fsf@m4x.org> (Alan Schmitt's message of "Fri, 19 Jun 2020 16:06:44 +0200") References: <87sger6tfv.fsf@m4x.org> Message-ID: Alan Schmitt writes: > I want to submit a paper to a conference that uses light double-blind > reviewing. A crucial part of the work described by the paper is a piece > of software, and I?feel that it would be difficult to assess the paper > without being able to run the software, simply to check that it does > what we claim (this is what a reviewer said in a previous attempt). So I > want to anonymize the code and include it as an artefact. But that means > I need to remove all the copyright info, which bothers me a little. > Should I just not worry about it, or are there better ways to share > artifacts anonymously? Submissions to journals and conferencs are generally confidential, and reviewers are required to treat them as such. I expect this would apply to software artifacts, too. If you are unsure about the policy of the conference you are submitting to, ask the program chair. Confidentiality would preclude your code without copyright notices from "escaping" to the public or being utilized by members of the program committee for purposes other than review of your submission. Eric. -- ------------------------------------------------------------------------------- Eric Eide . University of Utah School of Computing http://www.cs.utah.edu/~eeide/ . +1 (801) 585-5512 voice, +1 (801) 581-5843 FAX From gabriel.scherer at gmail.com Sat Jun 20 11:37:01 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Sat, 20 Jun 2020 17:37:01 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87sger6tfv.fsf@m4x.org> References: <87sger6tfv.fsf@m4x.org> Message-ID: If some piece of software does not come with copyright notices, it does *not* mean that it is in the public domain! (Just like: if you find a book manuscript or a painting without a signature, its copyright still belongs to its author.) By default, the usual copyright rules apply, and they are very restrictive (basically one is not allowed to do anything with the code or the software). Copyright notices establish authorship, but they are also crucial to *relax* those by-default restrictions by specifying a more permissive license to use the software or its source code. On Sat, Jun 20, 2020 at 3:10 PM Alan Schmitt wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Hello, > > I'm not sure this is the right place to ask this question, but I don't > know of a forum where I could reach as many researchers who might have > the same problem. Please tell me if there is a better place. > > I want to submit a paper to a conference that uses light double-blind > reviewing. A crucial part of the work described by the paper is a piece > of software, and I feel that it would be difficult to assess the paper > without being able to run the software, simply to check that it does > what we claim (this is what a reviewer said in a previous attempt). So I > want to anonymize the code and include it as an artefact. But that means > I need to remove all the copyright info, which bothers me a little. > Should I just not worry about it, or are there better ways to share > artifacts anonymously? > > Thanks, > > Alan > From Xavier.Leroy at inria.fr Sat Jun 20 12:10:23 2020 From: Xavier.Leroy at inria.fr (Xavier Leroy) Date: Sat, 20 Jun 2020 18:10:23 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: References: <87sger6tfv.fsf@m4x.org> Message-ID: On Sat, Jun 20, 2020 at 5:39 PM Gabriel Scherer wrote: > > If some piece of software does not come with copyright notices, it does > *not* mean that it is in the public domain! (Just like: if you find a book > manuscript or a painting without a signature, its copyright still belongs > to its author.) By default, the usual copyright rules apply, and they are > very restrictive (basically one is not allowed to do anything with the code > or the software). Exactly. And in case there was any doubt: copyright applies to anonymous or pseudonymous works, there is no requirement that the actual names of the authors are revealed. See https://commons.wikimedia.org/wiki/Commons:Anonymous_works . Regards, - Xavier Leroy > Copyright notices establish authorship, but they are also > crucial to *relax* those by-default restrictions by specifying a more > permissive license to use the software or its source code. > > On Sat, Jun 20, 2020 at 3:10 PM Alan Schmitt < > alan.schmitt at polytechnique.org> > wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > Hello, > > > > I'm not sure this is the right place to ask this question, but I don't > > know of a forum where I could reach as many researchers who might have > > the same problem. Please tell me if there is a better place. > > > > I want to submit a paper to a conference that uses light double-blind > > reviewing. A crucial part of the work described by the paper is a piece > > of software, and I feel that it would be difficult to assess the paper > > without being able to run the software, simply to check that it does > > what we claim (this is what a reviewer said in a previous attempt). So I > > want to anonymize the code and include it as an artefact. But that means > > I need to remove all the copyright info, which bothers me a little. > > Should I just not worry about it, or are there better ways to share > > artifacts anonymously? > > > > Thanks, > > > > Alan > > > From monnier at iro.umontreal.ca Sat Jun 20 13:21:25 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Sat, 20 Jun 2020 13:21:25 -0400 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87sger6tfv.fsf@m4x.org> (Alan Schmitt's message of "Fri, 19 Jun 2020 16:06:44 +0200") References: <87sger6tfv.fsf@m4x.org> Message-ID: > So I want to anonymize the code and include it as an artefact. > But that means I need to remove all the copyright info, which bothers > me a little. Should I just not worry about it, or are there better > ways to share artifacts anonymously? You should worry about it just as much (or just as little, as the case may be) as you worry about doing it for the paper itself ;-) Stefan From fjmd1a at gmail.com Sat Jun 20 14:06:07 2020 From: fjmd1a at gmail.com (Francis Davey) Date: Sun, 21 Jun 2020 03:06:07 +0900 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: References: <87sger6tfv.fsf@m4x.org> Message-ID: On Sun, Jun 21, 2020 at 1:42 AM Xavier Leroy wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Exactly. And in case there was any doubt: copyright applies to anonymous > or pseudonymous works, there is no requirement that the actual names of the > authors are revealed. See > https://commons.wikimedia.org/wiki/Commons:Anonymous_works . > > It is also worth saying that this is a fairly robust principle that pretty much all jurisdictions will apply this rule. It used to be the case that some jurisdictions (the USA being a late holdout) did require formalities of some kind for protection of copyright (in the USA registration still has some benefits, but I doubt it will matter to you). So the copyright should be yours from the moment you write it. There are some subtleties about anonymous works, in particular where the author cannot be found, but again I doubt they would be of any relevance to you. There idea that you need a (C) type statement persists though. (A long, long, time ago I did a PhD in theoretical computer science, but I am not an intellectual property lawyer, so I felt rather more confident contributing this answer). -- Francis Davey From fw at deneb.enyo.de Sun Jun 21 07:20:00 2020 From: fw at deneb.enyo.de (Florian Weimer) Date: Sun, 21 Jun 2020 13:20:00 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87sger6tfv.fsf@m4x.org> (Alan Schmitt's message of "Fri, 19 Jun 2020 16:06:44 +0200") References: <87sger6tfv.fsf@m4x.org> Message-ID: <87lfkg7jj3.fsf@mid.deneb.enyo.de> * Alan Schmitt: > I want to submit a paper to a conference that uses light double-blind > reviewing. A crucial part of the work described by the paper is a piece > of software, and I?feel that it would be difficult to assess the paper > without being able to run the software, simply to check that it does > what we claim (this is what a reviewer said in a previous attempt). So I > want to anonymize the code and include it as an artefact. But that means > I need to remove all the copyright info, which bothers me a little. > Should I just not worry about it, or are there better ways to share > artifacts anonymously? Is the software an original work created by the same authors as the paper? Then Stefan's comment applies. Otherwise, it's much more complicated. It definitely sounds problematic to remove copyright notices that you yourself did not add. On the other hand, people routinely do that when they compile software and upload it to someone else's computer for execution, without preserving these notices notices, even when the copyright notices explicitly state that distribution of binaries must preserve these notices. From tadeusz.litak at gmail.com Sun Jun 21 10:25:52 2020 From: tadeusz.litak at gmail.com (Tadeusz Litak) Date: Sun, 21 Jun 2020 16:25:52 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87lfkg7jj3.fsf@mid.deneb.enyo.de> References: <87sger6tfv.fsf@m4x.org> <87lfkg7jj3.fsf@mid.deneb.enyo.de> Message-ID: <26ae5940-68ba-ff75-7922-a874783699b7@gmail.com> On 21.06.20 13:20, Florian Weimer wrote: > Otherwise, it's much more complicated. It definitely sounds > problematic to remove copyright notices that you yourself did not add. > On the other hand, people routinely do that when they compile software > and upload it to someone else's computer for execution, without > preserving these notices notices, even when the copyright notices > explicitly state that distribution of binaries must preserve these > notices. Besides, if a generally available piece of software (not written by the authors of the submission) is involved, why would it be necessary to remove its corresponding copyright notices? In order to preserve anonymous character of the paper, isn't it enough to simply state that a given fragment of library involves third-party software, already in public domain and not necessarily written by submitting authors? Moreover, it is not even always necessary to include such code verbatim: one can use git submodules, git-lfs etc. It seems anyway that the original post was asking about the scenario where the software is entirely self-developed. But it is a good opportunity to clarify such issues in a broader context. From p.giarrusso at gmail.com Sun Jun 21 11:46:23 2020 From: p.giarrusso at gmail.com (Paolo Giarrusso) Date: Sun, 21 Jun 2020 17:46:23 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: <87lfkg7jj3.fsf@mid.deneb.enyo.de> References: <87sger6tfv.fsf@m4x.org> <87lfkg7jj3.fsf@mid.deneb.enyo.de> Message-ID: On Sun, 21 Jun 2020 at 15:52, Florian Weimer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > * Alan Schmitt: > > > I want to submit a paper to a conference that uses light double-blind > > reviewing. A crucial part of the work described by the paper is a piece > > of software, and I feel that it would be difficult to assess the paper > > without being able to run the software, simply to check that it does > > what we claim (this is what a reviewer said in a previous attempt). So I > > want to anonymize the code and include it as an artefact. But that means > > I need to remove all the copyright info, which bothers me a little. > > Should I just not worry about it, or are there better ways to share > > artifacts anonymously? > > Is the software an original work created by the same authors as the > paper? Then Stefan's comment applies. > Otherwise, it's much more complicated. It definitely sounds > problematic to remove copyright notices that you yourself did not add. > On the other hand, people routinely do that when they compile software > and upload it to someone else's computer for execution, without > preserving these notices notices, even when the copyright notices > explicitly state that distribution of binaries must preserve these > notices. But you need only remove names of the _authors_, not of anybody else, even if their name were a strong hint: light double-blind is not meant to _prevent_ inferring the names (at least in SIGPLAN's case). From alan.schmitt at polytechnique.org Sun Jun 21 14:39:07 2020 From: alan.schmitt at polytechnique.org (Alan Schmitt) Date: Sun, 21 Jun 2020 20:39:07 +0200 Subject: [TYPES] artifacts for double blind submissions: what about copyright? In-Reply-To: References: <87sger6tfv.fsf@m4x.org> <87lfkg7jj3.fsf@mid.deneb.enyo.de> Message-ID: <878sgg5kms.fsf@m4x.org> Good evening, Thanks everyone for the insightful answers. We'll go the easy route: a copy of the code with no copyright notice. I'm now convinced it's the simplest way. Thanks again, Alan From breuvart at lipn.univ-paris13.fr Mon Jul 27 12:19:50 2020 From: breuvart at lipn.univ-paris13.fr (Flavien Breuvart) Date: Mon, 27 Jul 2020 18:19:50 +0200 Subject: [TYPES] global debriefing over our virtual experience of conferences Message-ID: Dear colleagues, This spring, under unfortunate circumstances, many conferences held virtually. We have witnessed the disadvantages of such dispositives, but also its numerous advantages. Many of those conferences have had internal debates for debriefing this experiences, but I haven't seen any large and public debate inside the community. I was hopping that some of you may engage in such debates. As a starting point, I will try to succinctly expose my own point of view, which is probably subjective, politically charged, and highly debatable, but this is the whole point :-) I think we where all impressed by the high level of attendance of conferences and workshops. But when thinking back at it, this situation is perfectly normal as virtual conferences opened several blockades usually preventing people from coming, in particular via the absence of fees, the flexibility with respect to other duties (familial, teaching or administrative), or the weight of travels. Even if this was the only reason, I think it would be worth considering to secure part of these improvements. Another, huge (but politically charged) advantage, is the drastic reduction of the carbon footprint of our conferences. Several colleges are advocating for a public engagement of the community to reduce its global footprint. For example, see https://tcs4f.org/ which is a group advocating for a 50% carbon reduction in theoretical computer sciences. I have no doubt that other such initiative exist here and there; this year unfortunate event at least showed that they are well founded and not unreachable. That being said, I have to address the fact that our virtual conferences had technical issues and that physical ones have several other advantages. Concerning the technical issues (timeline clashes, internet connection, organization...), I strongly believe that time and experience can overcome most of them; I was helping in the organization comity of FSCD and it appear that many issues could have been avoided by a few technical adjustments (such as assigning two co-chairs for each sessions for example). Concerning the advantages of conferences, I see three important ones : 1) the chance encounters, 2) the strengthening of collaborations, and 3) the prolonged focus. 1) From my (short) experience, the first can happen in smaller scale meetings, that can be mostly local (with a minority of invited non-local visitors). 2) The best way to strengthen collaborations is not conferences but lab invitations (which could be more frequent without conferences fees and time expenditures). 3) I got the impression that most people where not as focus as in traditional conferences, but not to a big margin, and mainly by lack of routine (here I distinguish independent seminars and regular courses, as all teachers I have seen the disaster of virtualization among our students...). All in all, I would advocate for more small scale meetings, more lab invitations, but a virtualization of big scale conferences, and (why not), the securing of some international virtual seminar that where very interesting (thank you for the organizers that took those initiatives !). I hope I was not too long and too boring, do not hesitate to contradict me, all I want is to start a fruitful debate. Best, Flavien From nr at cs.tufts.edu Wed Jul 29 19:06:52 2020 From: nr at cs.tufts.edu (Norman Ramsey) Date: Wed, 29 Jul 2020 19:06:52 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? Message-ID: <20200729230652.89091903E50@homedog.cs.tufts.edu> This question might be off topic, but I hope not---there is a lot of knowledge here. Why is "lexical scoping" called "lexical"? I'm used to "lexical" referring to properties that describe how concrete syntax is formed---what constitutes a comment, how characters are grouped into tokens, and that sort of thing. The "lexical scope" used to give meaning to a closure seems more like a property of the syntax, not of the lexis. Is there another meaning of "lexical" that I don't know? Or some interesting history? Norman Ramsey From nikhil at acm.org Thu Jul 30 14:51:28 2020 From: nikhil at acm.org (Rishiyur Nikhil) Date: Thu, 30 Jul 2020 14:51:28 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: <20200729230652.89091903E50@homedog.cs.tufts.edu> References: <20200729230652.89091903E50@homedog.cs.tufts.edu> Message-ID: Isn't it just that the binding of a variable can be determined by looking at textually surrounding scopes ('lexical' from 'textual') as opposed to 'dynamic scoping' where a binding depends on the call chain and can refer to a binding that is textually arbitrary far away and unconnected with the text structure? Nikhil On Thu, Jul 30, 2020 at 1:27 PM Norman Ramsey wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > This question might be off topic, but I hope not---there is a lot of > knowledge here. Why is "lexical scoping" called "lexical"? > I'm used to "lexical" referring to properties that describe how > concrete syntax is formed---what constitutes a comment, how characters > are grouped into tokens, and that sort of thing. The "lexical scope" > used to give meaning to a closure seems more like a property of the > syntax, not of the lexis. > > Is there another meaning of "lexical" that I don't know? > Or some interesting history? > > > Norman Ramsey > > > From simone.martini at unibo.it Fri Jul 31 08:35:32 2020 From: simone.martini at unibo.it (Simone Martini) Date: Fri, 31 Jul 2020 12:35:32 +0000 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: <20200729230652.89091903E50@homedog.cs.tufts.edu> References: <20200729230652.89091903E50@homedog.cs.tufts.edu> Message-ID: Hi Norman Is there another meaning of "lexical" that I don't know? Or some interesting history? there could be people here who knows more than me, but I think the answer is ?no? on both count. First, I think that the expression "lexical scoping? is bad. In presence of the (more linguistically correct) use of the expression ?lexical analysis?, it seems to imply, as you suggest, that the two refer to the same notion of ?lexical?, which I think it is not the case. Both in research literature, and teaching, we should use instead ?static scoping?. BTW, most PL books use ?static?, not ?lexical?. A very quick review: de Bakker 1989; Pratt: 1985; Sethi: 1989: uses "lexical scope rule" (called also "static scope rule"); Appel book on compilers: 1998; Tucker and Noonan: 2002; Mitchell: 2003; Ghezzi&Jazayeri 1998 ?static binding (also called lexical scoping)?; Scott 2000, who in footnote says ?lexical scope is actually a better term than static scope, because scope rules based on nesting can be enforced at run time instead of compile time if desired?. I looked for clues in early papers. - Algol revised report, 1963 http://homepages.cs.ncl.ac.uk/cliff.jones/publications/OCRd/BBG63.pdf It uses the notion of ?scope?, but no use of the term ?lexical? (static scoping is implied by the ?copy rule?, which is used to explain the semantics of procedure calls). - Weizenbaum, The funarg problem explained, 1968 http://www.softwarepreservation.org/projects/LISP/MIT/Weizenbaum-FUNARG_Problem_Explained-1968.pdf no use of lexical - Gries, Compiler construction for digital computers, 1971 no use of lexical - Pascal report, 1972: no use of lexical Then, in Moses, The function of FUNCTION in Lisp, 1970 https://dl.acm.org/doi/10.1145/1093410.1093411 we find: ?the free variable _a_ in the program segment below is easily accessed because it is used lexically near the place where it is bound? (and again another occurrence in the same way) "Lexically near? is not incorrect (as it is instead ?lexical scope?), but it could have been here?and in this meaning (?closer in the text?) that the term ?lexical? starts its march? -s. PS I would be interested in pinpointing the exact place where the expression ?lexical scope? has been introduced. --------------------------------------------------- Simone Martini Dip. di Informatica - Scienza e Ingegneria Universita' di Bologna Mura Anteo Zamboni, 7 tel: +39 051 2094979 40127 Bologna BO www.cs.unibo.it/~martini Italy On 30 Jul 2020, at 01:06, Norman Ramsey > wrote: [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] This question might be off topic, but I hope not---there is a lot of knowledge here. Why is "lexical scoping" called "lexical"? I'm used to "lexical" referring to properties that describe how concrete syntax is formed---what constitutes a comment, how characters are grouped into tokens, and that sort of thing. The "lexical scope" used to give meaning to a closure seems more like a property of the syntax, not of the lexis. Is there another meaning of "lexical" that I don't know? Or some interesting history? Norman Ramsey From michael.greenberg at pomona.edu Fri Jul 31 12:49:24 2020 From: michael.greenberg at pomona.edu (Michael Greenberg) Date: Fri, 31 Jul 2020 09:49:24 -0700 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: <20200729230652.89091903E50@homedog.cs.tufts.edu> References: <20200729230652.89091903E50@homedog.cs.tufts.edu> Message-ID: Hi Norman, I don't have insight into the history, but a use of 'lexical' in the POSIX shell standard shows that the confusion between "lexical" and "syntactical" is pervasive and not just about binder scoping. The `break` special builtin utility has the conventional meaning: break from a loop. But consider the following program: ``` f() { break; } while true; do f; done ``` Should `f` break from the outer loop or not? The standard refers to that `break` as a "non-lexical break", using the phrase "lexically encloses" . (It turns out the behavior of non-lexically enclosing `break`s is unspecified, and shells do different things. Good luck!) I think "lexical" here (and in "lexical scoping") is meant to mean "pertaining to the [bracketing structure of the] written concrete syntax", even though that's not the dictionary meaning of the word "lexical" (as you point out). It would make more sense to say "syntactically enclosed", but... here we are. Cheers, Michael On 2020-07-29 19:06:52, Norman Ramsey wrote: > [ The Types Forum, https://nam01.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Cmichael.greenberg%40pomona.edu%7C06a319f5bd674be8f12908d834add6be%7C817f590439044ee8b3a5a65d4746ff70%7C0%7C0%7C637317268526469223&sdata=OuVmPhQajwWm68NyfOzdblzXmXBwTdh1PwD1gS9628g%3D&reserved=0 ] > > This question might be off topic, but I hope not---there is a lot of > knowledge here. Why is "lexical scoping" called "lexical"? > I'm used to "lexical" referring to properties that describe how > concrete syntax is formed---what constitutes a comment, how characters > are grouped into tokens, and that sort of thing. The "lexical scope" > used to give meaning to a closure seems more like a property of the > syntax, not of the lexis. > > Is there another meaning of "lexical" that I don't know? > Or some interesting history? > > > Norman Ramsey > > > ________________________________ > > [EXTERNAL EMAIL] Exercise caution before clicking on links or opening attachments. From jacob.errington at mail.mcgill.ca Fri Jul 31 08:56:42 2020 From: jacob.errington at mail.mcgill.ca (Jacob Thomas Errington) Date: Fri, 31 Jul 2020 08:56:42 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: <20200729230652.89091903E50@homedog.cs.tufts.edu> References: <20200729230652.89091903E50@homedog.cs.tufts.edu> Message-ID: <20200731125642.nvfr4z5ine3rmgoq@renro> Lexical scoping is also called static scoping. I believe it's called "lexical" because the variables in scope at a given point is determined by that point's position in the program text. In other words, the scoping is determined by the syntax. This is in contrast with dynamic scoping, where the set of variables in scope is determined by the execution of the program. Jake On Wed, Jul 29, 2020 at 07:06:52PM -0400, Norman Ramsey wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > This question might be off topic, but I hope not---there is a lot of > knowledge here. Why is "lexical scoping" called "lexical"? > I'm used to "lexical" referring to properties that describe how > concrete syntax is formed---what constitutes a comment, how characters > are grouped into tokens, and that sort of thing. The "lexical scope" > used to give meaning to a closure seems more like a property of the > syntax, not of the lexis. > > Is there another meaning of "lexical" that I don't know? > Or some interesting history? > > > Norman Ramsey > > -- Jacob Thomas Errington W: https://jerrington.me/ #: (514) 503-3100 From kbbruce47 at gmail.com Sat Aug 1 11:36:40 2020 From: kbbruce47 at gmail.com (Kim Bruce) Date: Sat, 1 Aug 2020 08:36:40 -0700 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: <20200729230652.89091903E50@homedog.cs.tufts.edu> References: <20200729230652.89091903E50@homedog.cs.tufts.edu> Message-ID: <691AF0B3-A293-4527-85DA-FD65A7315853@gmail.com> Interestingly, I grew up hearing ?static scoping? rather than ?lexical?. However, when learning about Smalltalk, started hearing ?lexical closure? for what most people these days simply refer to as ?closure?. Wikipedia (for what it is worth) gives the following history for the use of Closure: "The concept of closures was developed in the 1960s for the mechanical evaluation of expressions in the ?-calculus and was first fully implemented in 1970 as a language feature in the PAL programming language to support lexically scoped first-class functions .[2] :Turner's section 2, note 8 contains his claim about M-expressions Peter J. Landin defined the term closure in 1964 as having an environment part and a control part as used by his SECD machine for evaluating expressions.[3] Joel Moses credits Landin with introducing the term closure to refer to a lambda expression whose open bindings (free variables) have been closed by (or bound in) the lexical environment, resulting in a closed expression, or closure.[4] [5] This usage was subsequently adopted by Sussman and Steele when they defined Scheme in 1975,[6] a lexically scoped variant of LISP , and became widespread.? I didn?t track down who wrote this passage, but that may offer other hints as to its origin. Kim Bruce > On Jul 29, 2020, at 4:06 PM, Norman Ramsey wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > This question might be off topic, but I hope not---there is a lot of > knowledge here. Why is "lexical scoping" called "lexical"? > I'm used to "lexical" referring to properties that describe how > concrete syntax is formed---what constitutes a comment, how characters > are grouped into tokens, and that sort of thing. The "lexical scope" > used to give meaning to a closure seems more like a property of the > syntax, not of the lexis. > > Is there another meaning of "lexical" that I don't know? > Or some interesting history? > > > Norman Ramsey > > From selinger at mathstat.dal.ca Sat Aug 1 11:54:06 2020 From: selinger at mathstat.dal.ca (selinger at mathstat.dal.ca) Date: Sat, 1 Aug 2020 12:54:06 -0300 (ADT) Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: Message-ID: Michael Greenberg wrote: > > I think "lexical" here (and in "lexical scoping") is meant to mean > "pertaining to the [bracketing structure of the] written concrete > syntax", even though that's not the dictionary meaning of the word > "lexical" (as you point out). It would make more sense to say > "syntactically enclosed", but... here we are. Michael's reply prompted me to look up the actual dictionary meaning of the word "lexical", and I found two meanings (from Merriam-Webster). The first is: 1. "of or relating to words or the vocabulary of a language as distinguished from its grammar and construction" So that would be the meaning appropriate for phrases like "lexical analysis". But there's also: 2. "of or relating to a lexicon or to lexicography" That is the meaning I would associate with the phrase "lexical scoping". As I understand lexical scoping, a scope is a dictionary or lexicon, basically a map from identifiers of the concrete syntax to binding sites of the abstract syntax. This dictionary is updated during parsing. Perhaps "lexical" refers to the fact that the meaning of the identifier is looked up in this (static) lexicon, as opposed to determined dynamically at runtime? -- Peter From rwh at cs.cmu.edu Sun Aug 2 13:44:39 2020 From: rwh at cs.cmu.edu (Robert Harper) Date: Sun, 2 Aug 2020 13:44:39 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: References: Message-ID: <52530A93-0E0D-474C-8E98-25270E55CEB6@cs.cmu.edu> It seems to be overlooked that, often, it is not possible to perform identifier scope resolution during parsing (at least, not for what I take to be the standard meaning of the term ?parsing?). For example, in SML there is a form of declaration ?open S1. ? Sn? that brings into scope the identifiers determined by the type (ie, signature) of the path of structures. The parser can?t know this information; it is available only during type checking. In particular, it is not possible to examine the program text as such, find an identifier, and point to its binding site without also performing type checking. Although the terminology is not very precise, I would say that this is nevertheless ?static scope?, because type checking is done statically (without execution), but one may not want to call it ?lexical? (and so much the worse for that term, it being too simple-minded for practical purposes). Some time ago, as I recall, Francois Pottier wrote a paper studying a variety of exotic scoping concepts for PL?s, none of which would be ?lexical? in this simple sense. Again as a matter of terminology, I would argue that there is no such thing as ?dynamic scoping?. I think that is misleading terminology arising from confusing various things, such as classes, symbols, tags, and state, with scope. For example, even when dynamically allocated symbols have a definite static scope, though this can be obscured when their scope is ?extruded? (as they say in the pi calculus) to the outermost level. Bob (c) Robert Harper All Rights Reserved. > On Aug 1, 2020, at 11:54, selinger at mathstat.dal.ca wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Michael Greenberg wrote: >> >> I think "lexical" here (and in "lexical scoping") is meant to mean >> "pertaining to the [bracketing structure of the] written concrete >> syntax", even though that's not the dictionary meaning of the word >> "lexical" (as you point out). It would make more sense to say >> "syntactically enclosed", but... here we are. > > Michael's reply prompted me to look up the actual dictionary meaning > of the word "lexical", and I found two meanings (from > Merriam-Webster). The first is: > > 1. "of or relating to words or the vocabulary of a language as > distinguished from its grammar and construction" > > So that would be the meaning appropriate for phrases like "lexical > analysis". But there's also: > > 2. "of or relating to a lexicon or to lexicography" > > That is the meaning I would associate with the phrase "lexical > scoping". As I understand lexical scoping, a scope is a dictionary or > lexicon, basically a map from identifiers of the concrete syntax to > binding sites of the abstract syntax. This dictionary is updated > during parsing. Perhaps "lexical" refers to the fact that the meaning > of the identifier is looked up in this (static) lexicon, as opposed to > determined dynamically at runtime? > > -- Peter From monnier at iro.umontreal.ca Sun Aug 2 23:34:14 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Sun, 02 Aug 2020 23:34:14 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: (selinger@mathstat.dal.ca's message of "Sat, 1 Aug 2020 12:54:06 -0300 (ADT)") References: Message-ID: > 2. "of or relating to a lexicon or to lexicography" > > That is the meaning I would associate with the phrase "lexical > scoping". As I understand lexical scoping, a scope is a dictionary or > lexicon, basically a map from identifiers of the concrete syntax to > binding sites of the abstract syntax. This dictionary is updated > during parsing. Perhaps "lexical" refers to the fact that the meaning > of the identifier is looked up in this (static) lexicon, as opposed to > determined dynamically at runtime? I don't really buy that explanation: the implementation of dynamic scoping also uses a "dictionary" where variables are looked up. The rules for which dictionary to use when is just a bit different from those of static scoping, but the lookup of variables in a lexicon is common to both scoping rules. Stefan From nr at cs.tufts.edu Mon Aug 3 10:27:33 2020 From: nr at cs.tufts.edu (Norman Ramsey) Date: Mon, 03 Aug 2020 10:27:33 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: (sfid-H-20200802-194653-+123.07-1@multi.osbf.lua) References: (sfid-H-20200802-194653-+123.07-1@multi.osbf.lua) Message-ID: <20200803142733.73BFE1BC65FF@labrador.cs.tufts.edu> > 2. "of or relating to a lexicon or to lexicography" > > That is the meaning I would associate with the phrase "lexical > scoping". As I understand lexical scoping, a scope is a dictionary or > lexicon, basically a map from identifiers of the concrete syntax to > binding sites of the abstract syntax. This dictionary is updated > during parsing. Perhaps "lexical" refers to the fact that the meaning > of the identifier is looked up in this (static) lexicon, as opposed to > determined dynamically at runtime? That's a really interesting thought! I may politely ask Guy Steele what he remembers. Norman From hendrik at topoi.pooq.com Mon Aug 3 19:19:50 2020 From: hendrik at topoi.pooq.com (Hendrik Boom) Date: Mon, 3 Aug 2020 19:19:50 -0400 Subject: [TYPES] what is "lexical" about "lexical scoping"? In-Reply-To: References: Message-ID: <20200803231950.k23uxbadyj255o5t@topoi.pooq.com> On Sun, Aug 02, 2020 at 11:34:14PM -0400, Stefan Monnier wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > 2. "of or relating to a lexicon or to lexicography" > > > > That is the meaning I would associate with the phrase "lexical > > scoping". As I understand lexical scoping, a scope is a dictionary or > > lexicon, basically a map from identifiers of the concrete syntax to > > binding sites of the abstract syntax. This dictionary is updated > > during parsing. Perhaps "lexical" refers to the fact that the meaning > > of the identifier is looked up in this (static) lexicon, as opposed to > > determined dynamically at runtime? > > I don't really buy that explanation: the implementation of dynamic > scoping also uses a "dictionary" where variables are looked up. > The rules for which dictionary to use when is just a bit different from > those of static scoping, but the lookup of variables in a lexicon is > common to both scoping rules. Yes, both scoping rules use a lexicon. What makes dynamic scoping dynamic is that the lexicon is used at run time, and can even be different from one call of a function to another. I'm speaking the (in my mind execrable) habit of letting the variables available at the caller be the ones used in the callee. -- hendrik From fw at deneb.enyo.de Fri Aug 7 14:23:24 2020 From: fw at deneb.enyo.de (Florian Weimer) Date: Fri, 07 Aug 2020 20:23:24 +0200 Subject: [TYPES] what is "lexical" about "lexical scoping"? References: Message-ID: <87d0425n8j.fsf@mid.deneb.enyo.de> * Stefan Monnier: >> 2. "of or relating to a lexicon or to lexicography" >> >> That is the meaning I would associate with the phrase "lexical >> scoping". As I understand lexical scoping, a scope is a dictionary or >> lexicon, basically a map from identifiers of the concrete syntax to >> binding sites of the abstract syntax. This dictionary is updated >> during parsing. Perhaps "lexical" refers to the fact that the meaning >> of the identifier is looked up in this (static) lexicon, as opposed to >> determined dynamically at runtime? > > I don't really buy that explanation: the implementation of dynamic > scoping also uses a "dictionary" where variables are looked up. > The rules for which dictionary to use when is just a bit different from > those of static scoping, but the lookup of variables in a lexicon is > common to both scoping rules. I think there is an implementation strategy where dynamic scope has just a single, global environment, and shadowing of bindings is implemented by explicit saving and restoring of symbol values. One could argue that the saved values formed a dictionary of some sort, but it was not actually used for lookups, so that would be a stretch. With that in mind, maybe scopes were called lexical because they have lexicons associated with them, at least in a much more obvious way? From xialiyao at seas.upenn.edu Fri Aug 7 17:00:01 2020 From: xialiyao at seas.upenn.edu (Li-yao Xia) Date: Fri, 7 Aug 2020 17:00:01 -0400 Subject: [TYPES] Where does the name "algebraic data types" come from? Message-ID: <878b1211-fe8a-2477-1c2e-a7e2aa608eca@seas.upenn.edu> Dear TYPES, What makes "algebraic data types" algebraic? There are at least two common explanations: 1. Algebraic data types are defined using sums and products of types, by analogy with concepts from algebra. 2. Algebraic data types are initial algebras. Numerous blogs and papers explain "algebraic data types" that way, with a large majority adopting the more approachable explanation (1). The question is, how did the appellation "algebraic" come about historically? Did it start with one or both of the meanings above? The language feature, independently of the name, seems attributed to the HOPE language[1] (see for instance A History of Haskell: Being Lazy with Class[2]). However, the paper introducing HOPE does not use the word "algebraic" anywhere. Other papers from a cursory search on the topic do not discuss the origins of the word. I am curious about citations regarding the history of the name "algebraic data types" itself. Regards, Li-yao [1]: http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.18.8135 [2]: https://www.microsoft.com/en-us/research/wp-content/uploads/2016/07/history.pdf From serguei.lenglet at univ-lorraine.fr Tue Aug 18 12:06:22 2020 From: serguei.lenglet at univ-lorraine.fr (Serguei Lenglet) Date: Tue, 18 Aug 2020 18:06:22 +0200 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: References: Message-ID: Hello A. Donaldson published a detailed report on PLDI as its general chair https://www.doc.ic.ac.uk/~afd/homepages/papers/pdfs/2020/PLDIReport.pdf A very short personal summary: more people attended the virtual PLDI conference, but with less engagement (they attended less talks on average). The main issues are time zones and how to keep some form of impromptu meetings as in physical conferences. Maybe the organizers of other virtual conferences also published some feedback? Best regards Sergue? Le mar. 28 juil. 2020 ? 14:55, Flavien Breuvart < breuvart at lipn.univ-paris13.fr> a ?crit : > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear colleagues, > > This spring, under unfortunate circumstances, many conferences held > virtually. We have witnessed the disadvantages of such dispositives, but > also its numerous advantages. Many of those conferences have had > internal debates for debriefing this experiences, but I haven't seen any > large and public debate inside the community. I was hopping that some of > you may engage in such debates. > > As a starting point, I will try to succinctly expose my own point of > view, which is probably subjective, politically charged, and highly > debatable, but this is the whole point :-) > > I think we where all impressed by the high level of attendance of > conferences and workshops. But when thinking back at it, this situation > is perfectly normal as virtual conferences opened several blockades > usually preventing people from coming, in particular via the absence of > fees, the flexibility with respect to other duties (familial, teaching > or administrative), or the weight of travels. Even if this was the only > reason, I think it would be worth considering to secure part of these > improvements. > > Another, huge (but politically charged) advantage, is the drastic > reduction of the carbon footprint of our conferences. Several colleges > are advocating for a public engagement of the community to reduce its > global footprint. For example, see https://tcs4f.org/ which is a group > advocating for a 50% carbon reduction in theoretical computer sciences. > I have no doubt that other such initiative exist here and there; this > year unfortunate event at least showed that they are well founded and > not unreachable. > > That being said, I have to address the fact that our virtual conferences > had technical issues and that physical ones have several other > advantages. Concerning the technical issues (timeline clashes, internet > connection, organization...), I strongly believe that time and > experience can overcome most of them; I was helping in the organization > comity of FSCD and it appear that many issues could have been avoided by > a few technical adjustments (such as assigning two co-chairs for each > sessions for example). > > Concerning the advantages of conferences, I see three important ones : > 1) the chance encounters, 2) the strengthening of collaborations, and 3) > the prolonged focus. 1) From my (short) experience, the first can happen > in smaller scale meetings, that can be mostly local (with a minority of > invited non-local visitors). 2) The best way to strengthen > collaborations is not conferences but lab invitations (which could be > more frequent without conferences fees and time expenditures). 3) I got > the impression that most people where not as focus as in traditional > conferences, but not to a big margin, and mainly by lack of routine > (here I distinguish independent seminars and regular courses, as all > teachers I have seen the disaster of virtualization among our students...). > > All in all, I would advocate for more small scale meetings, more lab > invitations, but a virtualization of big scale conferences, and (why > not), the securing of some international virtual seminar that where very > interesting (thank you for the organizers that took those initiatives !). > > I hope I was not too long and too boring, do not hesitate to contradict > me, all I want is to start a fruitful debate. > > Best, > > Flavien > > > From gabriel.scherer at gmail.com Sun Aug 23 09:57:49 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Sun, 23 Aug 2020 15:57:49 +0200 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: References: Message-ID: I am in broad agreement with many of Flavien's points. I hope that we can learn how to do virtual conferences well so that we can reduce our travel footprint in the future, not just due to pandemic issues. In this respect I have been fairly impressed with the degree of investment of many members of our community in finding and building better tools for virtual conferences. Thanks! I hope that this major change (that is imposed to us for an unpredictable amount of time) could also be an occasion to seriously consider de-synchronizing publication of our work from conference presentations. I think that journal publications have better academic review process, but we've been traditionally tied to major conferences as publication venues. Maybe it is time to change this? In this respect an interesting approach is "The Art, Science, and Engineering of Programming" journal which is coupled with the conference: journal publishes four volumes a year (trying to fit a three-months reviewing process), and the conference is held annually, with all papers accepted during the year presented. Forced-online venues could be an occasion to experiment with this. (We could think of other formats, such as having a *seminar* attached to a journal instead of a conference; so far I found it easier to enjoy online seminars than online conferences.) On Tue, Jul 28, 2020 at 2:47 PM Flavien Breuvart < breuvart at lipn.univ-paris13.fr> wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear colleagues, > > This spring, under unfortunate circumstances, many conferences held > virtually. We have witnessed the disadvantages of such dispositives, but > also its numerous advantages. Many of those conferences have had > internal debates for debriefing this experiences, but I haven't seen any > large and public debate inside the community. I was hopping that some of > you may engage in such debates. > > As a starting point, I will try to succinctly expose my own point of > view, which is probably subjective, politically charged, and highly > debatable, but this is the whole point :-) > > I think we where all impressed by the high level of attendance of > conferences and workshops. But when thinking back at it, this situation > is perfectly normal as virtual conferences opened several blockades > usually preventing people from coming, in particular via the absence of > fees, the flexibility with respect to other duties (familial, teaching > or administrative), or the weight of travels. Even if this was the only > reason, I think it would be worth considering to secure part of these > improvements. > > Another, huge (but politically charged) advantage, is the drastic > reduction of the carbon footprint of our conferences. Several colleges > are advocating for a public engagement of the community to reduce its > global footprint. For example, see https://tcs4f.org/ which is a group > advocating for a 50% carbon reduction in theoretical computer sciences. > I have no doubt that other such initiative exist here and there; this > year unfortunate event at least showed that they are well founded and > not unreachable. > > That being said, I have to address the fact that our virtual conferences > had technical issues and that physical ones have several other > advantages. Concerning the technical issues (timeline clashes, internet > connection, organization...), I strongly believe that time and > experience can overcome most of them; I was helping in the organization > comity of FSCD and it appear that many issues could have been avoided by > a few technical adjustments (such as assigning two co-chairs for each > sessions for example). > > Concerning the advantages of conferences, I see three important ones : > 1) the chance encounters, 2) the strengthening of collaborations, and 3) > the prolonged focus. 1) From my (short) experience, the first can happen > in smaller scale meetings, that can be mostly local (with a minority of > invited non-local visitors). 2) The best way to strengthen > collaborations is not conferences but lab invitations (which could be > more frequent without conferences fees and time expenditures). 3) I got > the impression that most people where not as focus as in traditional > conferences, but not to a big margin, and mainly by lack of routine > (here I distinguish independent seminars and regular courses, as all > teachers I have seen the disaster of virtualization among our students...). > > All in all, I would advocate for more small scale meetings, more lab > invitations, but a virtualization of big scale conferences, and (why > not), the securing of some international virtual seminar that where very > interesting (thank you for the organizers that took those initiatives !). > > I hope I was not too long and too boring, do not hesitate to contradict > me, all I want is to start a fruitful debate. > > Best, > > Flavien > > > From gabriel.scherer at gmail.com Sun Aug 23 10:05:44 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Sun, 23 Aug 2020 16:05:44 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Dear types-list, Going on a tangent from Flavien's earlier post: I really think that online conferences should be free. Several conferences (PLDI for example) managed to run free-of-charge since the pandemic started, and they reported broader attendance and a strong diversity of attendants, which sounds great. I don't think we can achieve this with for-pay online conferences. ICFP is coming up shortly with a $100 registration price tag, and I did not register. I'm aware that running a large virtual conference requires computing resources that do have a cost. For PLDI for example, the report only says that the cost was covered by industrial sponsors. Are numbers publicly available on the cost of running a virtual conference? Note that if we managed to run a conference on free software, I'm sure that institutions and volunteers could be convinced to help hosting and monitoring the conference services during the event. From tringer at cs.washington.edu Sun Aug 23 10:14:12 2020 From: tringer at cs.washington.edu (Talia Ringer) Date: Sun, 23 Aug 2020 07:14:12 -0700 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: I don't know about PLDI, but there are some costs associated with online events. For example, automatic captioning software is still not very good (Google's always turns "proofs" into "fruits" for me). Live captioning is really expensive! But it's also hugely important for disability accessibility. For students, ICFP was essentially free. I do agree that in principle, online conferences should be free, and online components of hybrid conferences should be free or strongly discounted. In practice, though, I do think that will mean finding sponsors for hidden costs that really are necessary. On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear types-list, > > Going on a tangent from Flavien's earlier post: I really think that online > conferences should be free. > > Several conferences (PLDI for example) managed to run free-of-charge since > the pandemic started, and they reported broader attendance and a strong > diversity of attendants, which sounds great. I don't think we can achieve > this with for-pay online conferences. > > ICFP is coming up shortly with a $100 registration price tag, and I did not > register. > > I'm aware that running a large virtual conference requires computing > resources that do have a cost. For PLDI for example, the report only says > that the cost was covered by industrial sponsors. Are numbers publicly > available on the cost of running a virtual conference? Note that if we > managed to run a conference on free software, I'm sure that institutions > and volunteers could be convinced to help hosting and monitoring the > conference services during the event. > From mwh at cs.umd.edu Sun Aug 23 10:53:41 2020 From: mwh at cs.umd.edu (Michael Hicks) Date: Sun, 23 Aug 2020 10:53:41 -0400 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Thanks for raising this issue. Just a few points about the other side of the argument: It?s well known that things that are free are not valued (by humans) as much as those that cost something, even a small amount. For example, see Dan Ariely?s ?Predictably Irrational? which presents the results of several experiments that demonstrate this. As a relevant case: Free MOOCs tended to have lots of ?sign ups? but far fewer attendees, and even fewer completers. As such, if the goal is to have engaged attendees, trying to come closer to the experience of traditional conferences, it might make sense to charge something, even a small amount like $25, for at least some of the population. This population might be people who have lots of social capital already, and are generally busy, so they are more likely to blow off the conference if they paid nothing for signing up. Such people might be those that more junior attendees wish to meet. I note that engaged attendance was a goal when we had in-person conferences, so I don?t see why we?d want to drop it now. Indeed, if people don?t want to be engaged the videos will be available for free, afterward. Beyond the modest fees to run an online conference, which Talia mentions, conference registration payments serve other purposes. Any surplus goes to SIGPLAN, which turns around this surplus as good works, e.g., paying the open access fees for PACMPL, which ICFP benefits from. It also makes donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. Corporate sponsors can indeed pay some costs, but they also have downsides. We are finding that many sponsors are not interested in necessarily giving that much, and some are starting to make demands on how the conference is run for their modest donation. This is a slippery slope that the SIGPLAN EC is trying to avoid. Given that PLDI was completely free and ICFP followed a progressive fee schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, to see how the registration fee affected attendance. Thanks, -Mike On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > > > I don't know about PLDI, but there are some costs associated with online > > events. For example, automatic captioning software is still not very good > > (Google's always turns "proofs" into "fruits" for me). Live captioning is > > really expensive! But it's also hugely important for disability > > accessibility. > > > > For students, ICFP was essentially free. I do agree that in principle, > > online conferences should be free, and online components of hybrid > > conferences should be free or strongly discounted. In practice, though, I > > do think that will mean finding sponsors for hidden costs that really are > > necessary. > > > > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer > > > wrote: > > > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > ] > > > > > > Dear types-list, > > > > > > Going on a tangent from Flavien's earlier post: I really think that > online > > > conferences should be free. > > > > > > Several conferences (PLDI for example) managed to run free-of-charge > since > > > the pandemic started, and they reported broader attendance and a strong > > > diversity of attendants, which sounds great. I don't think we can achieve > > > this with for-pay online conferences. > > > > > > ICFP is coming up shortly with a $100 registration price tag, and I did > not > > > register. > > > > > > I'm aware that running a large virtual conference requires computing > > > resources that do have a cost. For PLDI for example, the report only says > > > that the cost was covered by industrial sponsors. Are numbers publicly > > > available on the cost of running a virtual conference? Note that if we > > > managed to run a conference on free software, I'm sure that institutions > > > and volunteers could be convinced to help hosting and monitoring the > > > conference services during the event. > > > > > From nicolai.kraus at gmail.com Sun Aug 23 12:35:34 2020 From: nicolai.kraus at gmail.com (Nicolai Kraus) Date: Sun, 23 Aug 2020 17:35:34 +0100 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Interesting discussion, and definitely very important! My opinion is: (1) Registration costs should not stop anyone from attending a/an [online] conference. I guess that's obvious and solutions for this were implemented for physical conferences. (2) I accept Mike's point about free things not being valued as highly as paid things. But I think even a small symbolic fee could potentially be a hurdle for some people. The issue is that the value of $25 (or ?25 or ?25 or whatever) is very subjective. For those senior people who are important for the conference and who are the ones that junior members want to meet, $25 is likely to be negligible. For the junior participants, it might not be. This is just the wrong way round since the junior participants probably benefit most from the meeting and don't need this sort of encouragement. Of course, the perceived value of $25 will also greatly depend on whether someone has access to academic travel budget. Finally, we shouldn't forget that a significant part of the world population (online sources say 25%, no idea how accurate this is) has no access to a bank account which makes even a fee of $0.01 a problem. Someone with this background could not attend a physical conference, but they might have access to the internet. I don't know whether we will actually have such participants, but we (we = the privileged inhabitants of developed countries) would be ignorant if we dismissed the possibility. (3) I'm against relying on industrial sponsors. How much advertisement at conferences is acceptable? It's hard to draw a line, and this could get out of hand. Moreover, this route of funding might not be available for some more theory-focussed conferences, and I assume it would in general benefit large/prestigious conferences much more than small/new meetings. (4) I actually liked the model that LICS used. Participants could choose between free registration and paid registration, with the condition that each paper came with one paid registration to cover the publication costs. I believe we could instead simply say that people with access to travel budget are kindly asked to opt for the paid registration. I do think that this would quite easily cover the costs for the conference. Best, Nicolai On Sun, Aug 23, 2020 at 3:58 PM Michael Hicks wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Thanks for raising this issue. Just a few points about the other side of > the argument: > > It?s well known that things that are free are not valued (by humans) as > much as those that cost something, even a small amount. For example, see > Dan Ariely?s ?Predictably Irrational? which presents the results of several > experiments that demonstrate this. As a relevant case: Free MOOCs tended to > have lots of ?sign ups? but far fewer attendees, and even fewer completers. > > As such, if the goal is to have engaged attendees, trying to come closer to > the experience of traditional conferences, it might make sense to charge > something, even a small amount like $25, for at least some of the > population. This population might be people who have lots of social capital > already, and are generally busy, so they are more likely to blow off the > conference if they paid nothing for signing up. Such people might be those > that more junior attendees wish to meet. > > I note that engaged attendance was a goal when we had in-person > conferences, so I don?t see why we?d want to drop it now. Indeed, if people > don?t want to be engaged the videos will be available for free, afterward. > > Beyond the modest fees to run an online conference, which Talia mentions, > conference registration payments serve other purposes. Any surplus goes to > SIGPLAN, which turns around this surplus as good works, e.g., paying the > open access fees for PACMPL, which ICFP benefits from. It also makes > donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. > > Corporate sponsors can indeed pay some costs, but they also have downsides. > We are finding that many sponsors are not interested in necessarily giving > that much, and some are starting to make demands on how the conference is > run for their modest donation. This is a slippery slope that the SIGPLAN EC > is trying to avoid. > > Given that PLDI was completely free and ICFP followed a progressive fee > schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, > to see how the registration fee affected attendance. > > Thanks, > -Mike > > On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer > wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > > > > > I don't know about PLDI, but there are some costs associated with online > > > > events. For example, automatic captioning software is still not very good > > > > (Google's always turns "proofs" into "fruits" for me). Live captioning is > > > > really expensive! But it's also hugely important for disability > > > > accessibility. > > > > > > > > For students, ICFP was essentially free. I do agree that in principle, > > > > online conferences should be free, and online components of hybrid > > > > conferences should be free or strongly discounted. In practice, though, I > > > > do think that will mean finding sponsors for hidden costs that really are > > > > necessary. > > > > > > > > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < > gabriel.scherer at gmail.com > > > > > > > wrote: > > > > > > > > > [ The Types Forum, > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > > > ] > > > > > > > > > > Dear types-list, > > > > > > > > > > Going on a tangent from Flavien's earlier post: I really think that > > online > > > > > conferences should be free. > > > > > > > > > > Several conferences (PLDI for example) managed to run free-of-charge > > since > > > > > the pandemic started, and they reported broader attendance and a strong > > > > > diversity of attendants, which sounds great. I don't think we can > achieve > > > > > this with for-pay online conferences. > > > > > > > > > > ICFP is coming up shortly with a $100 registration price tag, and I did > > not > > > > > register. > > > > > > > > > > I'm aware that running a large virtual conference requires computing > > > > > resources that do have a cost. For PLDI for example, the report only > says > > > > > that the cost was covered by industrial sponsors. Are numbers publicly > > > > > available on the cost of running a virtual conference? Note that if we > > > > > managed to run a conference on free software, I'm sure that > institutions > > > > > and volunteers could be convinced to help hosting and monitoring the > > > > > conference services during the event. > > > > > > > > > > From tringer at cs.washington.edu Sun Aug 23 11:11:21 2020 From: tringer at cs.washington.edu (Talia Ringer) Date: Sun, 23 Aug 2020 08:11:21 -0700 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: References: Message-ID: I used to argue against changing conference deadline systems a lot, but the pandemic response and political events in the US have made it clear to me that this is a diversity issue. When a deadline is only once per year (and some of us do not have work that easily crosses over to other major conferences), missing it can be a major setback. And events like the pandemic have a disproportionate impact on groups that are traditionally underrepresented in our field. So the consequences of the deadline system are very uneven and reinforce our field's current demographic. I agree that it is absolutely prudent to take this opportunity to reflect on our review process. I want to push strongly for moving to a model in which deadlines exist but are much more frequent (say, monthly or quarterly). I think the approach Gabriel Scherer mentioned that is taken by "The Art, Science, and Engineering of Programming" would be better both for science and for diversity in our field. Talia https://dependenttyp.es/ On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > I am in broad agreement with many of Flavien's points. I hope that we can > learn how to do virtual conferences well so that we can reduce our travel > footprint in the future, not just due to pandemic issues. In this respect I > have been fairly impressed with the degree of investment of many members of > our community in finding and building better tools for virtual conferences. > Thanks! > > I hope that this major change (that is imposed to us for an unpredictable > amount of time) could also be an occasion to seriously consider > de-synchronizing publication of our work from conference presentations. I > think that journal publications have better academic review process, but > we've been traditionally tied to major conferences as publication venues. > Maybe it is time to change this? In this respect an interesting approach is > "The Art, Science, and Engineering of Programming" journal which is coupled > with the conference: journal publishes four volumes a year > (trying to fit a three-months reviewing process), and the conference is > held annually, with all papers accepted during the year presented. > Forced-online venues could be an occasion to experiment with this. (We > could think of other formats, such as having a *seminar* attached to a > journal instead of a conference; so far I found it easier to enjoy online > seminars than online conferences.) > > > On Tue, Jul 28, 2020 at 2:47 PM Flavien Breuvart < > breuvart at lipn.univ-paris13.fr> wrote: > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > Dear colleagues, > > > > This spring, under unfortunate circumstances, many conferences held > > virtually. We have witnessed the disadvantages of such dispositives, but > > also its numerous advantages. Many of those conferences have had > > internal debates for debriefing this experiences, but I haven't seen any > > large and public debate inside the community. I was hopping that some of > > you may engage in such debates. > > > > As a starting point, I will try to succinctly expose my own point of > > view, which is probably subjective, politically charged, and highly > > debatable, but this is the whole point :-) > > > > I think we where all impressed by the high level of attendance of > > conferences and workshops. But when thinking back at it, this situation > > is perfectly normal as virtual conferences opened several blockades > > usually preventing people from coming, in particular via the absence of > > fees, the flexibility with respect to other duties (familial, teaching > > or administrative), or the weight of travels. Even if this was the only > > reason, I think it would be worth considering to secure part of these > > improvements. > > > > Another, huge (but politically charged) advantage, is the drastic > > reduction of the carbon footprint of our conferences. Several colleges > > are advocating for a public engagement of the community to reduce its > > global footprint. For example, see https://tcs4f.org/ which is a group > > advocating for a 50% carbon reduction in theoretical computer sciences. > > I have no doubt that other such initiative exist here and there; this > > year unfortunate event at least showed that they are well founded and > > not unreachable. > > > > That being said, I have to address the fact that our virtual conferences > > had technical issues and that physical ones have several other > > advantages. Concerning the technical issues (timeline clashes, internet > > connection, organization...), I strongly believe that time and > > experience can overcome most of them; I was helping in the organization > > comity of FSCD and it appear that many issues could have been avoided by > > a few technical adjustments (such as assigning two co-chairs for each > > sessions for example). > > > > Concerning the advantages of conferences, I see three important ones : > > 1) the chance encounters, 2) the strengthening of collaborations, and 3) > > the prolonged focus. 1) From my (short) experience, the first can happen > > in smaller scale meetings, that can be mostly local (with a minority of > > invited non-local visitors). 2) The best way to strengthen > > collaborations is not conferences but lab invitations (which could be > > more frequent without conferences fees and time expenditures). 3) I got > > the impression that most people where not as focus as in traditional > > conferences, but not to a big margin, and mainly by lack of routine > > (here I distinguish independent seminars and regular courses, as all > > teachers I have seen the disaster of virtualization among our > students...). > > > > All in all, I would advocate for more small scale meetings, more lab > > invitations, but a virtualization of big scale conferences, and (why > > not), the securing of some international virtual seminar that where very > > interesting (thank you for the organizers that took those initiatives !). > > > > I hope I was not too long and too boring, do not hesitate to contradict > > me, all I want is to start a fruitful debate. > > > > Best, > > > > Flavien > > > > > > > From nicolai.kraus at gmail.com Sun Aug 23 12:42:35 2020 From: nicolai.kraus at gmail.com (Nicolai Kraus) Date: Sun, 23 Aug 2020 17:42:35 +0100 Subject: [TYPES] Fwd: Panel Debate, Wednesday 2 September @ 3pm UTC: "Evolution or Revolution? The Future of Conferences in Theoretical Computer Science" In-Reply-To: References: Message-ID: Hi all, I'd like to advertise the below panel debate on the future of TCS conferences. The discussion thread "online conferences should be free" would also be relevant for this debate. Cheers, Nicolai ---------- Forwarded message --------- From: Jamie Vicary Date: Mon, Aug 17, 2020 at 10:54 PM Subject: Panel Debate, Wednesday 2 September @ 3pm UTC: "Evolution or Revolution? The Future of Conferences in Theoretical Computer Science" To: [TLDR: Panel Debate on the Future of Conferences in TCS -- panel Fong, Kesner, Pierce, Vardi -- join at 3pm UTC on Weds 2 Sep -- zoom.us/j/177472153 -- reply now from academic email address to propose a question -- please circulate this message widely] Dear all, The entire community is invited to participate in a debate on the future of the conference system in theoretical computer science. Organized as a special event as part of the Online Worldwide Seminar on Logic and Semantics (OWLS), this will provide a rare community-wide opportunity for us to consider the strengths and weaknesses of our current system, and consider if we can do better. The scope of the debate is all aspects of our publishing and community traditions, characterised by prestige earned mostly through publication in competitive conferences, and frequent local and international travel. Possible topics for discussion include the need to publish in conferences for career progression, which usually involves burning carbon; wasted author and reviewer effort when good papers are rejected from highly competitive conferences; the extent of our responsibility as a community to respond to climate change; alternative publishing models, like the journal-focussed system used in mathematics; high costs of conference travel and registration; virtual conference advantages, disadvantages and best practice; improving equality, diversity and access; consequences and response to COVID-19; and the role of professional bodies. These topics have many close relationships, and need to be discussed together to gain a full understanding of the issues involved, and how we can move forward. OUR PANEL To discuss these issues, we have an excellent panel with a wide range of relevant experience: - Dr Brendan Fong, MIT (brendanfong.com) is a postdoctoral researcher with considerable experience organizing virtual conferences and seminars (act2020.mit.edu), and an Executive Editor of the new open-access journal Compositionality. - Professor Delia Kesner, University of Paris (irif.fr/~kesner) has served on the Steering Committee of six conferences and workshops, and is currently the SC Chair of FSCD, the most recent iteration of which was organized at short notice as a virtual event (fscd2020.org). - Professor Benjamin Pierce, University of Pennsylvania (cis.upenn.edu/~bcpierce) has served as PC chair of a range of events including POPL and ICFP, and has written powerfully on the need for the computer science community to adapt to the reality of climate change. - Professor Moshe Vardi, Rice University (cs.rice.edu/~vardi) is Senior Editor of the journal Communications of the ACM, and founded the Federated Logic Conference (FLOC). He has long been a vocal commentator on structural problems with computer science publishing. PROPOSE A QUESTION Questions will be asked by members of the community. That means you! Please reply to this email to propose your question, which could raise any issue in scope. **Why not do it right now?** Make sure to use an academic email address. We'll let you know if your question is accepted, and you'll then have the opportunity to ask it during the debate, and to respond to the panel's comments. WHEN AND WHERE The debate will take place on Wednesday 2 September at 3pm UTC, which corresponds to the following times in a range of cities around the world: 8am San Francisco / 10am Houston / 11am Philadelphia / 4pm London / 5pm Paris / 9pm Mumbai / 11pm Beijing / midnight Tokyo / 1am Sydney The event will take place on Zoom at the following address, with no password or registration required: - zoom.us/j/177472153 The debate will be followed by an opportunity to discuss informally with other members of the community in small groups. WEBPAGE This event is organized as part of the OWLS seminar series. For more information, a calendar you can embed into your own, and to sign up for reminder emails, visit the webpage: - https://www.cs.bham.ac.uk/~vicaryjo/owls/ READING Members of the community may enjoy the following articles, related to the topic of the debate. - Lance Fortnow (2009), "Time for Computer Science to Grow Up", https://cacm.acm.org/magazines/2009/8/34492-viewpoint-time-for-computer-science-to-grow-up - Benjamin Pierce, Michael Hicks, Crista Lopes and Jens Palsberg (2020), "Conferences in an Era of Expensive Carbon", https://cacm.acm.org/magazines/2020/3/243024-conferences-in-an-era-of-expensive-carbon - Moshe Vardi (2020), "Publish and Perish", https://cacm.acm.org/magazines/2020/1/241717-publish-and-perish We hope you will join us for the debate. Please forward this message to members of your research group, and others who may be interested to participate. Best wishes, Jamie From h.basold at liacs.leidenuniv.nl Sun Aug 23 15:23:15 2020 From: h.basold at liacs.leidenuniv.nl (Henning Basold) Date: Sun, 23 Aug 2020 21:23:15 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: I would like to add another way of covering costs that is often used in communal places: Anyone gives whatever they can, which may be nothing. This assumes of course some fairness and some transparency about the costs. The problem with the LICS model is that it also prevents publication for some people, if they cannot collaborate with someone who has money. That things with a higher price are perceived to be to of higher quality is unfortunately true. But do we have to reproduce this kind of marketing within our scientific community? Lastly, I would like to also mention the excellent journal LMCS (logical methods in CS), which has a very strong board and rolling deadlines. This journal implements many of the suggestions already successfully. On 23 August 2020 18:35:34 CEST, Nicolai Kraus wrote: >[ The Types Forum, >http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > >Interesting discussion, and definitely very important! >My opinion is: > >(1) Registration costs should not stop anyone from attending a/an >[online] >conference. I guess that's obvious and solutions for this were >implemented >for physical conferences. > >(2) I accept Mike's point about free things not being valued as highly >as >paid things. But I think even a small symbolic fee could potentially be >a >hurdle for some people. The issue is that the value of $25 (or ?25 or >?25 >or whatever) is very subjective. For those senior people who are >important >for the conference and who are the ones that junior members want to >meet, >$25 is likely to be negligible. For the junior participants, it might >not >be. This is just the wrong way round since the junior participants >probably >benefit most from the meeting and don't need this sort of >encouragement. Of >course, the perceived value of $25 will also greatly depend on whether >someone has access to academic travel budget. Finally, we shouldn't >forget >that a significant part of the world population (online sources say >25%, no >idea how accurate this is) has no access to a bank account which makes >even >a fee of $0.01 a problem. Someone with this background could not attend >a >physical conference, but they might have access to the internet. I >don't >know whether we will actually have such participants, but we (we = the >privileged inhabitants of developed countries) would be ignorant if we >dismissed the possibility. > >(3) I'm against relying on industrial sponsors. How much advertisement >at >conferences is acceptable? It's hard to draw a line, and this could get >out >of hand. Moreover, this route of funding might not be available for >some >more theory-focussed conferences, and I assume it would in general >benefit >large/prestigious conferences much more than small/new meetings. > >(4) I actually liked the model that LICS used. Participants could >choose >between free registration and paid registration, with the condition >that >each paper came with one paid registration to cover the publication >costs. >I believe we could instead simply say that people with access to travel >budget are kindly asked to opt for the paid registration. I do think >that >this would quite easily cover the costs for the conference. > >Best, >Nicolai > >On Sun, Aug 23, 2020 at 3:58 PM Michael Hicks wrote: > >> [ The Types Forum, >http://lists.seas.upenn.edu/mailman/listinfo/types-list >> ] >> >> Thanks for raising this issue. Just a few points about the other side >of >> the argument: >> >> It?s well known that things that are free are not valued (by humans) >as >> much as those that cost something, even a small amount. For example, >see >> Dan Ariely?s ?Predictably Irrational? which presents the results of >several >> experiments that demonstrate this. As a relevant case: Free MOOCs >tended to >> have lots of ?sign ups? but far fewer attendees, and even fewer >completers. >> >> As such, if the goal is to have engaged attendees, trying to come >closer to >> the experience of traditional conferences, it might make sense to >charge >> something, even a small amount like $25, for at least some of the >> population. This population might be people who have lots of social >capital >> already, and are generally busy, so they are more likely to blow off >the >> conference if they paid nothing for signing up. Such people might be >those >> that more junior attendees wish to meet. >> >> I note that engaged attendance was a goal when we had in-person >> conferences, so I don?t see why we?d want to drop it now. Indeed, if >people >> don?t want to be engaged the videos will be available for free, >afterward. >> >> Beyond the modest fees to run an online conference, which Talia >mentions, >> conference registration payments serve other purposes. Any surplus >goes to >> SIGPLAN, which turns around this surplus as good works, e.g., paying >the >> open access fees for PACMPL, which ICFP benefits from. It also makes >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >> >> Corporate sponsors can indeed pay some costs, but they also have >downsides. >> We are finding that many sponsors are not interested in necessarily >giving >> that much, and some are starting to make demands on how the >conference is >> run for their modest donation. This is a slippery slope that the >SIGPLAN EC >> is trying to avoid. >> >> Given that PLDI was completely free and ICFP followed a progressive >fee >> schedule, I?ll be curious to compare the ICFP outbrief with that of >PLDI?s, >> to see how the registration fee affected attendance. >> >> Thanks, >> -Mike >> >> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer > >> wrote: >> >> > [ The Types Forum, >> http://lists.seas.upenn.edu/mailman/listinfo/types-list >> > ] >> > >> > >> > >> > I don't know about PLDI, but there are some costs associated with >online >> > >> > events. For example, automatic captioning software is still not >very good >> > >> > (Google's always turns "proofs" into "fruits" for me). Live >captioning is >> > >> > really expensive! But it's also hugely important for disability >> > >> > accessibility. >> > >> > >> > >> > For students, ICFP was essentially free. I do agree that in >principle, >> > >> > online conferences should be free, and online components of hybrid >> > >> > conferences should be free or strongly discounted. In practice, >though, I >> > >> > do think that will mean finding sponsors for hidden costs that >really are >> > >> > necessary. >> > >> > >> > >> > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >> gabriel.scherer at gmail.com >> > > >> > >> > wrote: >> > >> > >> > >> > > [ The Types Forum, >> > http://lists.seas.upenn.edu/mailman/listinfo/types-list >> > >> > > ] >> > >> > > >> > >> > > Dear types-list, >> > >> > > >> > >> > > Going on a tangent from Flavien's earlier post: I really think >that >> > online >> > >> > > conferences should be free. >> > >> > > >> > >> > > Several conferences (PLDI for example) managed to run >free-of-charge >> > since >> > >> > > the pandemic started, and they reported broader attendance and a >strong >> > >> > > diversity of attendants, which sounds great. I don't think we can >> achieve >> > >> > > this with for-pay online conferences. >> > >> > > >> > >> > > ICFP is coming up shortly with a $100 registration price tag, and >I did >> > not >> > >> > > register. >> > >> > > >> > >> > > I'm aware that running a large virtual conference requires >computing >> > >> > > resources that do have a cost. For PLDI for example, the report >only >> says >> > >> > > that the cost was covered by industrial sponsors. Are numbers >publicly >> > >> > > available on the cost of running a virtual conference? Note that >if we >> > >> > > managed to run a conference on free software, I'm sure that >> institutions >> > >> > > and volunteers could be convinced to help hosting and monitoring >the >> > >> > > conference services during the event. >> > >> > > >> > >> > >> From nicolai.kraus at gmail.com Sun Aug 23 20:14:07 2020 From: nicolai.kraus at gmail.com (Nicolai Kraus) Date: Mon, 24 Aug 2020 01:14:07 +0100 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: On Sun, Aug 23, 2020 at 8:23 PM Henning Basold wrote: > I would like to add another way of covering costs that is often used in > communal places: Anyone gives whatever they can, which may be nothing. This > assumes of course some fairness and some transparency about the costs. > "Pay what you want/can" is in many situations a nice model and I'm confident that transparency/fairness wouldn't be an issue in this community. But the problem is that conference fees are in most cases paid from grants and not privately. Grant holders have some responsibility to spend their budget in a way that benefits the funded project most, and in many cases, it's essentially taxpayers' money. Asking people to be generous with essentially public money would give rise to all sorts of ethical problems and actual questions. As an example, I've read in the rules of a funding agency that you can't pay for CO2 offset if that's a voluntary option when you buy your flight ticket. I first thought that this was an unfortunate rule, but it actually makes sense: This public money was allocated to research. Had the government (or whoever is responsible) thought that the money was better spent on CO2 reduction, then they could have done that instead. As a grant holder, one just isn't entitled to spend the money on environmental protection (even if that might benefit the public more than the actual research project). I guess it's debatable, but I don't think "pay what you want" works here. However, it does work to say "registration is ?200, but you can also choose free registration if you don't have a grant/access to travel budget or are in a similar situation; choose at your discretion." > The problem with the LICS model is that it also prevents publication for > some people, if they cannot collaborate with someone who has money. > Agreed, but this is not in any way a new problem. I think the LICS model was very reasonable and pragmatic given the circumstances. That doesn't mean that we can't come up with an even better solution if we discuss this as a community. > That things with a higher price are perceived to be to of higher quality > is unfortunately true. But do we have to reproduce this kind of marketing > within our scientific community? > I don't think "reproduce" is the right word, it's an unfortunate reality which we can take into account. Or we can choose to ignore this reality and hope that it works. I think it could work. > Lastly, I would like to also mention the excellent journal LMCS (logical > methods in CS), which has a very strong board and rolling deadlines. This > journal implements many of the suggestions already successfully. > Thanks for this comment! There will also be a public discussion on the publication culture in our communities (I've forwarded the invitation to the panel debate to this mailing list earlier today). Best wishes Nicolai > > On 23 August 2020 18:35:34 CEST, Nicolai Kraus > wrote: >> >> [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >> >> Interesting discussion, and definitely very important! >> My opinion is: >> >> (1) Registration costs should not stop anyone from attending a/an [online] >> conference. I guess that's obvious and solutions for this were implemented >> for physical conferences. >> >> (2) I accept Mike's point about free things not being valued as highly as >> paid things. But I think even a small symbolic fee could potentially be a >> hurdle for some people. The issue is that the value of $25 (or ?25 or ?25 >> or whatever) is very subjective. For those senior people who are important >> for the conference and who are the ones that junior members want to meet, >> $25 is likely to be negligible. For the junior participants, it might not >> be. This is just the wrong way round since the junior participants probably >> benefit most from the meeting and don't need this sort of encouragement. Of >> course, the perceived value of $25 will also greatly depend on whether >> someone has access to academic travel budget. Finally, we shouldn't forget >> that a significant part of the world population (online sources say 25%, no >> idea how accurate this is) has no access to a bank account which makes even >> a fee of $0.01 a problem. Someone with this background could not attend a >> physical conference, but they might have access to the internet. I don't >> know whether we will actually have such participants, but we (we = the >> privileged inhabitants of developed countries) would be ignorant if we >> dismissed the possibility. >> >> (3) I'm against relying on industrial sponsors. How much advertisement at >> conferences is acceptable? It's hard to draw a line, and this could get out >> of hand. Moreover, this route of funding might not be available for some >> more theory-focussed conferences, and I assume it would in general benefit >> large/prestigious conferences much more than small/new meetings. >> >> (4) I actually liked the model that LICS used. Participants could choose >> between free registration and paid registration, with the condition that >> each paper came with one paid registration to cover the publication costs. >> I believe we could instead simply say that people with access to travel >> budget are kindly asked to opt for the paid registration. I do think that >> this would quite easily cover the costs for the conference. >> >> Best, >> Nicolai >> >> On Sun, Aug 23, 2020 at 3:58 PM Michael Hicks wrote: >> >> [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list >>> ] >>> >>> Thanks for raising this issue. Just a few points about the other side of >>> the argument: >>> >>> It?s well known that things that are free are not valued (by humans) as >>> much as those that cost something, even a small amount. For example, see >>> Dan Ariely?s ?Predictably Irrational? which presents the results of several >>> experiments that demonstrate this. As a relevant case: Free MOOCs tended to >>> have lots of ?sign ups? but far fewer attendees, and even fewer completers. >>> >>> As such, if the goal is to have engaged attendees, trying to come closer to >>> the experience of traditional conferences, it might make sense to charge >>> something, even a small amount like $25, for at least some of the >>> population. This population might be people who have lots of social capital >>> already, and are generally busy, so they are more likely to blow off the >>> conference if they paid nothing for signing up. Such people might be those >>> that more junior attendees wish to meet. >>> >>> I note that engaged attendance was a goal when we had in-person >>> conferences, so I don?t see why we?d want to drop it now. Indeed, if people >>> don?t want to be engaged the videos will be available for free, afterward. >>> >>> Beyond the modest fees to run an online conference, which Talia mentions, >>> conference registration payments serve other purposes. Any surplus goes to >>> SIGPLAN, which turns around this surplus as good works, e.g., paying the >>> open access fees for PACMPL, which ICFP benefits from. It also makes >>> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >>> >>> Corporate sponsors can indeed pay some costs, but they also have downsides. >>> We are finding that many sponsors are not interested in necessarily giving >>> that much, and some are starting to make demands on how the conference is >>> run for their modest donation. This is a slippery slope that the SIGPLAN EC >>> is trying to avoid. >>> >>> Given that PLDI was completely free and ICFP followed a progressive fee >>> schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, >>> to see how the registration fee affected attendance. >>> >>> Thanks, >>> -Mike >>> >>> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer >>> wrote: >>> >>> [ The Types Forum, >>>> >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>> >>>> ] >>>> >>>> >>>> >>>> I don't know about PLDI, but there are some costs associated with online >>>> >>>> events. For example, automatic captioning software is still not very good >>>> >>>> (Google's always turns "proofs" into "fruits" for me). Live captioning is >>>> >>>> really expensive! But it's also hugely important for disability >>>> >>>> accessibility. >>>> >>>> >>>> >>>> For students, ICFP was essentially free. I do agree that in principle, >>>> >>>> online conferences should be free, and online components of hybrid >>>> >>>> conferences should be free or strongly discounted. In practice, though, I >>>> >>>> do think that will mean finding sponsors for hidden costs that really are >>>> >>>> necessary. >>>> >>>> >>>> >>>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >>>> >>> gabriel.scherer at gmail.com >>> >>>> wrote: >>>> >>>> >>>> >>>> [ The Types Forum, >>>>> >>>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>>> >>>> ] >>>>> Dear types-list, >>>>> Going on a tangent from Flavien's earlier post: I really think that >>>>> >>>> online >>>> >>>> conferences should be free. >>>>> Several conferences (PLDI for example) managed to run free-of-charge >>>>> >>>> since >>>> >>>> the pandemic started, and they reported broader attendance and a strong >>>>> diversity of attendants, which sounds great. I don't think we can >>>>> >>>> achieve >>> >>>> this with for-pay online conferences. >>>>> ICFP is coming up shortly with a $100 registration price tag, and I did >>>>> >>>> not >>>> >>>> register. >>>>> I'm aware that running a large virtual conference requires computing >>>>> resources that do have a cost. For PLDI for example, the report only >>>>> >>>> says >>> >>>> that the cost was covered by industrial sponsors. Are numbers publicly >>>>> available on the cost of running a virtual conference? Note that if we >>>>> managed to run a conference on free software, I'm sure that >>>>> >>>> institutions >>> >>>> and volunteers could be convinced to help hosting and monitoring the >>>>> conference services during the event. >>>>> >>>> >>>> From jon at jonmsterling.com Sun Aug 23 20:33:30 2020 From: jon at jonmsterling.com (Jon Sterling) Date: Sun, 23 Aug 2020 20:33:30 -0400 Subject: [TYPES] =?utf-8?q?global_debriefing_over_our_virtual_experience_?= =?utf-8?q?of_conferences?= In-Reply-To: References: Message-ID: <96f32b5b-5522-4f66-837b-6ca3bfc20db4@www.fastmail.com> I want to put in my agreement with Talia's second point that we should move to a model in which deadlines are frequent and cheap to miss (as is the case in every discipline that is lucky enough to be based on journals) --- I would add that if we can move toward such a system, it would probably be unnecessary to argue for the deadline extensions that have such a detrimental effect on the work-life balance of scientists (pace Talia's first point, which is well taken). I've missed submitting to POPL several times because the deadlines didn't line up with the stage of my research, but I am dreaming of a future where it really matters less for me that my research happens to be "medium rare" on approximately July 1st each year. I also support and agree with everything that Gabriel has said in this thread. I truly love workshops and seminars, and if I could flush all these expensive and stressful conferences directly down the toilet (together with the dogmatic ideology of our professional organizations and their representatives, which we consume en gavage) and instead just go to lovely low-stress workshops and send my work periodically to journals, I would be so happy. Best, Jon On Sun, Aug 23, 2020, at 11:11 AM, Talia Ringer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > I used to argue against changing conference deadline systems a lot, but the > pandemic response and political events in the US have made it clear to me > that this is a diversity issue. When a deadline is only once per year (and > some of us do not have work that easily crosses over to other major > conferences), missing it can be a major setback. And events like the > pandemic have a disproportionate impact on groups that are traditionally > underrepresented in our field. So the consequences of the deadline system > are very uneven and reinforce our field's current demographic. > > I agree that it is absolutely prudent to take this opportunity to reflect > on our review process. I want to push strongly for moving to a model in > which deadlines exist but are much more frequent (say, monthly or > quarterly). I think the approach Gabriel Scherer mentioned that is taken by > "The Art, Science, and Engineering of Programming" would be better both for > science and for diversity in our field. > > Talia > https://dependenttyp.es/ > > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer > wrote: > > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > I am in broad agreement with many of Flavien's points. I hope that we can > > learn how to do virtual conferences well so that we can reduce our travel > > footprint in the future, not just due to pandemic issues. In this respect I > > have been fairly impressed with the degree of investment of many members of > > our community in finding and building better tools for virtual conferences. > > Thanks! > > > > I hope that this major change (that is imposed to us for an unpredictable > > amount of time) could also be an occasion to seriously consider > > de-synchronizing publication of our work from conference presentations. I > > think that journal publications have better academic review process, but > > we've been traditionally tied to major conferences as publication venues. > > Maybe it is time to change this? In this respect an interesting approach is > > "The Art, Science, and Engineering of Programming" journal which is coupled > > with the conference: journal publishes four volumes a year > > (trying to fit a three-months reviewing process), and the conference is > > held annually, with all papers accepted during the year presented. > > Forced-online venues could be an occasion to experiment with this. (We > > could think of other formats, such as having a *seminar* attached to a > > journal instead of a conference; so far I found it easier to enjoy online > > seminars than online conferences.) > > > > > > On Tue, Jul 28, 2020 at 2:47 PM Flavien Breuvart < > > breuvart at lipn.univ-paris13.fr> wrote: > > > > > [ The Types Forum, > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > ] > > > > > > Dear colleagues, > > > > > > This spring, under unfortunate circumstances, many conferences held > > > virtually. We have witnessed the disadvantages of such dispositives, but > > > also its numerous advantages. Many of those conferences have had > > > internal debates for debriefing this experiences, but I haven't seen any > > > large and public debate inside the community. I was hopping that some of > > > you may engage in such debates. > > > > > > As a starting point, I will try to succinctly expose my own point of > > > view, which is probably subjective, politically charged, and highly > > > debatable, but this is the whole point :-) > > > > > > I think we where all impressed by the high level of attendance of > > > conferences and workshops. But when thinking back at it, this situation > > > is perfectly normal as virtual conferences opened several blockades > > > usually preventing people from coming, in particular via the absence of > > > fees, the flexibility with respect to other duties (familial, teaching > > > or administrative), or the weight of travels. Even if this was the only > > > reason, I think it would be worth considering to secure part of these > > > improvements. > > > > > > Another, huge (but politically charged) advantage, is the drastic > > > reduction of the carbon footprint of our conferences. Several colleges > > > are advocating for a public engagement of the community to reduce its > > > global footprint. For example, see https://tcs4f.org/ which is a group > > > advocating for a 50% carbon reduction in theoretical computer sciences. > > > I have no doubt that other such initiative exist here and there; this > > > year unfortunate event at least showed that they are well founded and > > > not unreachable. > > > > > > That being said, I have to address the fact that our virtual conferences > > > had technical issues and that physical ones have several other > > > advantages. Concerning the technical issues (timeline clashes, internet > > > connection, organization...), I strongly believe that time and > > > experience can overcome most of them; I was helping in the organization > > > comity of FSCD and it appear that many issues could have been avoided by > > > a few technical adjustments (such as assigning two co-chairs for each > > > sessions for example). > > > > > > Concerning the advantages of conferences, I see three important ones : > > > 1) the chance encounters, 2) the strengthening of collaborations, and 3) > > > the prolonged focus. 1) From my (short) experience, the first can happen > > > in smaller scale meetings, that can be mostly local (with a minority of > > > invited non-local visitors). 2) The best way to strengthen > > > collaborations is not conferences but lab invitations (which could be > > > more frequent without conferences fees and time expenditures). 3) I got > > > the impression that most people where not as focus as in traditional > > > conferences, but not to a big margin, and mainly by lack of routine > > > (here I distinguish independent seminars and regular courses, as all > > > teachers I have seen the disaster of virtualization among our > > students...). > > > > > > All in all, I would advocate for more small scale meetings, more lab > > > invitations, but a virtualization of big scale conferences, and (why > > > not), the securing of some international virtual seminar that where very > > > interesting (thank you for the organizers that took those initiatives !). > > > > > > I hope I was not too long and too boring, do not hesitate to contradict > > > me, all I want is to start a fruitful debate. > > > > > > Best, > > > > > > Flavien > > > > > > > > > > > > From tringer at cs.washington.edu Mon Aug 24 02:07:13 2020 From: tringer at cs.washington.edu (Talia Ringer) Date: Sun, 23 Aug 2020 23:07:13 -0700 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: <96f32b5b-5522-4f66-837b-6ca3bfc20db4@www.fastmail.com> References: <96f32b5b-5522-4f66-837b-6ca3bfc20db4@www.fastmail.com> Message-ID: I also think the annual cycle is self-reinforcing. If you miss it one year (say, thanks to a pandemic), you may submit your work to a smaller venue (say, CPP) shortly after. Then you start the next project later, and probably the next year you find yourself in the same situation all over again. In grad school, that cycle is really difficult to break. On workshops versus conferences, one potential downside of smaller workshops over conferences is that students from universities with few PL faculty (or students in industry) really need connections, and conferences maximize for connections. Changing the need for connections would be a very difficult and long process, and the right substitution is not clear. But regardless of what we do, we must make sure that we offer at least some equally good way for students without those connections to make them. And for me that was walking around at POPL during my first year and introducing myself to everyone I saw even if I was terrified and had no idea what I was doing. On Sun, Aug 23, 2020, 10:22 PM Jon Sterling wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > I want to put in my agreement with Talia's second point that we should > move to a model in which deadlines are frequent and cheap to miss (as is > the case in every discipline that is lucky enough to be based on journals) > --- I would add that if we can move toward such a system, it would probably > be unnecessary to argue for the deadline extensions that have such a > detrimental effect on the work-life balance of scientists (pace Talia's > first point, which is well taken). > > I've missed submitting to POPL several times because the deadlines didn't > line up with the stage of my research, but I am dreaming of a future where > it really matters less for me that my research happens to be "medium rare" > on approximately July 1st each year. > > I also support and agree with everything that Gabriel has said in this > thread. I truly love workshops and seminars, and if I could flush all these > expensive and stressful conferences directly down the toilet (together with > the dogmatic ideology of our professional organizations and their > representatives, which we consume en gavage) and instead just go to lovely > low-stress workshops and send my work periodically to journals, I would be > so happy. > > Best, > Jon > > > On Sun, Aug 23, 2020, at 11:11 AM, Talia Ringer wrote: > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > > > I used to argue against changing conference deadline systems a lot, but > the > > pandemic response and political events in the US have made it clear to me > > that this is a diversity issue. When a deadline is only once per year > (and > > some of us do not have work that easily crosses over to other major > > conferences), missing it can be a major setback. And events like the > > pandemic have a disproportionate impact on groups that are traditionally > > underrepresented in our field. So the consequences of the deadline system > > are very uneven and reinforce our field's current demographic. > > > > I agree that it is absolutely prudent to take this opportunity to reflect > > on our review process. I want to push strongly for moving to a model in > > which deadlines exist but are much more frequent (say, monthly or > > quarterly). I think the approach Gabriel Scherer mentioned that is taken > by > > "The Art, Science, and Engineering of Programming" would be better both > for > > science and for diversity in our field. > > > > Talia > > https://dependenttyp.es/ > > > > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < > gabriel.scherer at gmail.com> > > wrote: > > > > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > ] > > > > > > I am in broad agreement with many of Flavien's points. I hope that we > can > > > learn how to do virtual conferences well so that we can reduce our > travel > > > footprint in the future, not just due to pandemic issues. In this > respect I > > > have been fairly impressed with the degree of investment of many > members of > > > our community in finding and building better tools for virtual > conferences. > > > Thanks! > > > > > > I hope that this major change (that is imposed to us for an > unpredictable > > > amount of time) could also be an occasion to seriously consider > > > de-synchronizing publication of our work from conference > presentations. I > > > think that journal publications have better academic review process, > but > > > we've been traditionally tied to major conferences as publication > venues. > > > Maybe it is time to change this? In this respect an interesting > approach is > > > "The Art, Science, and Engineering of Programming" journal which is > coupled > > > with the conference: journal publishes four volumes a > year > > > (trying to fit a three-months reviewing process), and the conference is > > > held annually, with all papers accepted during the year presented. > > > Forced-online venues could be an occasion to experiment with this. (We > > > could think of other formats, such as having a *seminar* attached to a > > > journal instead of a conference; so far I found it easier to enjoy > online > > > seminars than online conferences.) > > > > > > > > > On Tue, Jul 28, 2020 at 2:47 PM Flavien Breuvart < > > > breuvart at lipn.univ-paris13.fr> wrote: > > > > > > > [ The Types Forum, > > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > > > ] > > > > > > > > Dear colleagues, > > > > > > > > This spring, under unfortunate circumstances, many conferences held > > > > virtually. We have witnessed the disadvantages of such dispositives, > but > > > > also its numerous advantages. Many of those conferences have had > > > > internal debates for debriefing this experiences, but I haven't seen > any > > > > large and public debate inside the community. I was hopping that > some of > > > > you may engage in such debates. > > > > > > > > As a starting point, I will try to succinctly expose my own point of > > > > view, which is probably subjective, politically charged, and highly > > > > debatable, but this is the whole point :-) > > > > > > > > I think we where all impressed by the high level of attendance of > > > > conferences and workshops. But when thinking back at it, this > situation > > > > is perfectly normal as virtual conferences opened several blockades > > > > usually preventing people from coming, in particular via the absence > of > > > > fees, the flexibility with respect to other duties (familial, > teaching > > > > or administrative), or the weight of travels. Even if this was the > only > > > > reason, I think it would be worth considering to secure part of these > > > > improvements. > > > > > > > > Another, huge (but politically charged) advantage, is the drastic > > > > reduction of the carbon footprint of our conferences. Several > colleges > > > > are advocating for a public engagement of the community to reduce its > > > > global footprint. For example, see https://tcs4f.org/ which is a > group > > > > advocating for a 50% carbon reduction in theoretical computer > sciences. > > > > I have no doubt that other such initiative exist here and there; this > > > > year unfortunate event at least showed that they are well founded and > > > > not unreachable. > > > > > > > > That being said, I have to address the fact that our virtual > conferences > > > > had technical issues and that physical ones have several other > > > > advantages. Concerning the technical issues (timeline clashes, > internet > > > > connection, organization...), I strongly believe that time and > > > > experience can overcome most of them; I was helping in the > organization > > > > comity of FSCD and it appear that many issues could have been > avoided by > > > > a few technical adjustments (such as assigning two co-chairs for each > > > > sessions for example). > > > > > > > > Concerning the advantages of conferences, I see three important ones > : > > > > 1) the chance encounters, 2) the strengthening of collaborations, > and 3) > > > > the prolonged focus. 1) From my (short) experience, the first can > happen > > > > in smaller scale meetings, that can be mostly local (with a minority > of > > > > invited non-local visitors). 2) The best way to strengthen > > > > collaborations is not conferences but lab invitations (which could be > > > > more frequent without conferences fees and time expenditures). 3) I > got > > > > the impression that most people where not as focus as in traditional > > > > conferences, but not to a big margin, and mainly by lack of routine > > > > (here I distinguish independent seminars and regular courses, as all > > > > teachers I have seen the disaster of virtualization among our > > > students...). > > > > > > > > All in all, I would advocate for more small scale meetings, more lab > > > > invitations, but a virtualization of big scale conferences, and (why > > > > not), the securing of some international virtual seminar that where > very > > > > interesting (thank you for the organizers that took those > initiatives !). > > > > > > > > I hope I was not too long and too boring, do not hesitate to > contradict > > > > me, all I want is to start a fruitful debate. > > > > > > > > Best, > > > > > > > > Flavien > > > > > > > > > > > > > > > > > > From h.basold at liacs.leidenuniv.nl Mon Aug 24 02:48:59 2020 From: h.basold at liacs.leidenuniv.nl (Henning Basold) Date: Mon, 24 Aug 2020 08:48:59 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> On 24/08/2020 02:14, Nicolai Kraus wrote: > On Sun, Aug 23, 2020 at 8:23 PM Henning Basold > > wrote: > > I would like to add another way of covering costs that is often used > in communal places: Anyone gives whatever they can, which may be > nothing. This assumes of course some fairness and some transparency > about the costs. > > > "Pay what you want/can" is in many situations a nice model and I'm > confident that transparency/fairness wouldn't be an issue in this > community. But the problem is that conference fees are in most cases > paid from grants and not privately. Grant holders have some > responsibility to spend their budget in a way that benefits the funded > project most, and in many cases, it's essentially taxpayers' money. > Asking people to be generous with essentially public money would give > rise to all sorts of ethical problems and actual questions.? > As an example, I've read in the rules of a funding agency that you can't > pay for CO2 offset if that's a voluntary option when you buy your flight > ticket. I first thought that this was an unfortunate rule, but it > actually makes sense: This public money was allocated to research. Had > the government (or whoever is responsible) thought that the money was > better spent on CO2 reduction, then they could have done that instead. > As a grant holder, one just isn't entitled to spend the money on > environmental protection (even if that might benefit the public more > than the actual research project). > I guess it's debatable, but I don't think "pay what you want" works > here. However, it does work to say "registration is??200, but you can > also choose free registration if you don't have a grant/access to travel > budget or are in a similar?situation; choose at your discretion." > ? This is indeed an issue with our funding model and the potential ethical issues could be resolved, if there was any political will; but let us leave this for another discussion. I think the model could be refined a bit from pay all or nothing to a pay scale, with the option of no payment, that may be linked with some criteria that allow the registrant to justify the selection to their funding agency. I think, we could, as a community, come up with some standard criteria that can employed by event organisers. > > The problem with the LICS model is that it also prevents publication > for some people, if they cannot collaborate with someone who has money. > > > Agreed, but this is not in any way a new problem. I think the LICS model > was very reasonable and pragmatic given the circumstances. That doesn't > mean that we can't come up with an even better solution if we discuss > this as a community. > ? Certainly not, and it has become ever more visible with paid open access, where publishers asked for astronomical fees. And yes, I think we can, and need to, come up with better solutions. > > That things with a higher price are perceived to be to of higher > quality is unfortunately true. But do we have to reproduce this kind > of marketing within our scientific community? > > > I don't think "reproduce" is the right word, it's an unfortunate reality > which we can take into account. Or we can choose to ignore this reality > and hope that it works. I think it could work. > ? I use the word "reproduce" purposefully because this kind of reality descends merely from our society and is not an absolute truth. Many scientists do value free publications platforms very highly, and often more so than certain paid journals. The key here is the community network that surrounds such a platform. It is, therefore, not the scientific community that is an issue but whatever institutions at whose grace they are for positions, pay cheques, funding etc. Fortunately, there are changes happening, like the DORA agreement (https://sfdora.org/). Changing the mindset from "quality must be expensive" to other criteria will also contribute to changing other issue, like the peer reviewing process at competitive conferences and the issue of competition and patriarchy itself, which have been discussed previously in this thread. > > Lastly, I would like to also mention the excellent journal LMCS > (logical methods in CS), which has a very strong board and rolling > deadlines. This journal implements many of the suggestions already > successfully. > > > Thanks for this comment! There will also be a public discussion on the > publication culture in our communities (I've forwarded the invitation to > the panel debate to this mailing list earlier today). > I'm looking forward to this discussion, and I hope that we can find, as a community, solutions to pressing issues and move forward. Best, Henning > Best wishes > Nicolai > > ? > > > On 23 August 2020 18:35:34 CEST, Nicolai Kraus > > wrote: > > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Interesting discussion, and definitely very important! > My opinion is: > > (1) Registration costs should not stop anyone from attending a/an [online] > conference. I guess that's obvious and solutions for this were implemented > for physical conferences. > > (2) I accept Mike's point about free things not being valued as highly as > paid things. But I think even a small symbolic fee could potentially be a > hurdle for some people. The issue is that the value of $25 (or ?25 or ?25 > or whatever) is very subjective. For those senior people who are important > for the conference and who are the ones that junior members want to meet, > $25 is likely to be negligible. For the junior participants, it might not > be. This is just the wrong way round since the junior participants probably > benefit most from the meeting and don't need this sort of encouragement. Of > course, the perceived value of $25 will also greatly depend on whether > someone has access to academic travel budget. Finally, we shouldn't forget > that a significant part of the world population (online sources say 25%, no > idea how accurate this is) has no access to a bank account which makes even > a fee of $0.01 a problem. Someone with this background could not attend a > physical conference, but they might have access to the internet. I don't > know whether we will actually have such participants, but we (we = the > privileged inhabitants of developed countries) would be ignorant if we > dismissed the possibility. > > (3) I'm against relying on industrial sponsors. How much advertisement at > conferences is acceptable? It's hard to draw a line, and this could get out > of hand. Moreover, this route of funding might not be available for some > more theory-focussed conferences, and I assume it would in general benefit > large/prestigious conferences much more than small/new meetings. > > (4) I actually liked the model that LICS used. Participants could choose > between free registration and paid registration, with the condition that > each paper came with one paid registration to cover the publication costs. > I believe we could instead simply say that people with access to travel > budget are kindly asked to opt for the paid registration. I do think that > this would quite easily cover the costs for the conference. > > Best, > Nicolai > > On Sun, Aug 23, 2020 at 3:58 PM Michael Hicks > wrote: > > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Thanks for raising this issue. Just a few points about the > other side of > the argument: > > It?s well known that things that are free are not valued (by > humans) as > much as those that cost something, even a small amount. For > example, see > Dan Ariely?s ?Predictably Irrational? which presents the > results of several > experiments that demonstrate this. As a relevant case: Free > MOOCs tended to > have lots of ?sign ups? but far fewer attendees, and even > fewer completers. > > As such, if the goal is to have engaged attendees, trying to > come closer to > the experience of traditional conferences, it might make > sense to charge > something, even a small amount like $25, for at least some > of the > population. This population might be people who have lots of > social capital > already, and are generally busy, so they are more likely to > blow off the > conference if they paid nothing for signing up. Such people > might be those > that more junior attendees wish to meet. > > I note that engaged attendance was a goal when we had in-person > conferences, so I don?t see why we?d want to drop it now. > Indeed, if people > don?t want to be engaged the videos will be available for > free, afterward. > > Beyond the modest fees to run an online conference, which > Talia mentions, > conference registration payments serve other purposes. Any > surplus goes to > SIGPLAN, which turns around this surplus as good works, > e.g., paying the > open access fees for PACMPL, which ICFP benefits from. It > also makes > donations to CRA-W, OPLSS, etc. and provides scholarships > for PLMW. > > Corporate sponsors can indeed pay some costs, but they also > have downsides. > We are finding that many sponsors are not interested in > necessarily giving > that much, and some are starting to make demands on how the > conference is > run for their modest donation. This is a slippery slope that > the SIGPLAN EC > is trying to avoid. > > Given that PLDI was completely free and ICFP followed a > progressive fee > schedule, I?ll be curious to compare the ICFP outbrief with > that of PLDI?s, > to see how the registration fee affected attendance. > > Thanks, > -Mike > > On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer > > > wrote: > > [ The Types Forum, > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > > > > I don't know about PLDI, but there are some costs > associated with online > > events. For example, automatic captioning software is > still not very good > > (Google's always turns "proofs" into "fruits" for me). > Live captioning is > > really expensive! But it's also hugely important for > disability > > accessibility. > > > > For students, ICFP was essentially free. I do agree that > in principle, > > online conferences should be free, and online components > of hybrid > > conferences should be free or strongly discounted. In > practice, though, I > > do think that will mean finding sponsors for hidden > costs that really are > > necessary. > > > > On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < > > gabriel.scherer at gmail.com > > wrote: > > > > [ The Types Forum, > > http://lists.seas.upenn.edu/mailman/listinfo/types-list > > ] > Dear types-list, > Going on a tangent from Flavien's earlier post: I > really think that > > online > > conferences should be free. > Several conferences (PLDI for example) managed to > run free-of-charge > > since > > the pandemic started, and they reported broader > attendance and a strong > diversity of attendants, which sounds great. I don't > think we can > > achieve > > this with for-pay online conferences. > ICFP is coming up shortly with a $100 registration > price tag, and I did > > not > > register. > I'm aware that running a large virtual conference > requires computing > resources that do have a cost. For PLDI for example, > the report only > > says > > that the cost was covered by industrial sponsors. > Are numbers publicly > available on the cost of running a virtual > conference? Note that if we > managed to run a conference on free software, I'm > sure that > > institutions > > and volunteers could be convinced to help hosting > and monitoring the > conference services during the event. > > From Buday.Gergely.Istvan at szie.hu Mon Aug 24 03:09:47 2020 From: Buday.Gergely.Istvan at szie.hu (Buday Gergely =?utf-8?b?SXN0dsOhbg==?=) Date: Mon, 24 Aug 2020 09:09:47 +0200 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: <96f32b5b-5522-4f66-837b-6ca3bfc20db4@www.fastmail.com> References: <96f32b5b-5522-4f66-837b-6ca3bfc20db4@www.fastmail.com> Message-ID: <20200824090947.Horde.EywYkUKwefCI4SbGWXq817O@webmail.ih.szie.hu> Related to this, I quote from my unpublished manuscript: This year the conferences are mostly held online. In the computer science research community we should discuss the question whether organizers should relax the requirement of physically attending conferences if you wanted to publish. The business model of journals is a heated debate, so could be the conferences. Why build a paywall for publications that does not go to reviewers but to the hotel, catering and travel industry? In poor regions going to expensive conferences is a problem. The drawbacks of online conferences are obvious, but the advantages are also. Why attend a conference if you are interested only in a section and you can follow it through online broadcast? ? Interestingly, even before we thought anything about a possible pandemic, there were voices of not holding that many conferences, due to the concern of the ecologic footprint of long-distance flights \cite{Vardi-Publish-and-Perish}. https://doi.org/10.1145/3373386 A response on twitter says ? \begin{quote}It's a noble goal to reduce the CO${}_2$ footprint. However, if you made physical attendance optional, neither universities nor companies would support conference participation. A decrease in physical attendance would also likely make conferences financially infeasible. \cite{Leidner-declining-conferences} \end{quote} ? I disagree here: conferences might morph into something we haven't seen yet, new business models can arise but the need for scientific discussion and publications will keep them alive. Online conferences let isolated researchers to participate and possibly joining distributed research groups. ? An interesting outcome would be the merge of journals and conferences, as it is exemplified by the Proceedings of the ACM on Programming Languages, which published papers from premier conferences like the Principles of Programming Languages (POPL). ? A conference held online this year, Logic in Computer Science already changed how we do a conference: participants are expected to watch presentations recorded in advance, letting more time to discuss research at videoconference time. ? A reviewer raised my attention about the drawbacks of moving conferences online: it is more difficult to build new friendships online, it is harder to get ideas from hallway discussions, no matter there are hallway Zoom channels for us. Another viewpoint is that funding organizations would spend less on conferences and so scientic life would be less human and more mechanized, as other parts of our life. Less discovering the world, less reward for spending much time in front of your laptop. That reviewer finds this effect on our culture risky. --- - Gergely Id?zet (Jon Sterling ): > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > I want to put in my agreement with Talia's second point that we > should move to a model in which deadlines are frequent and cheap to > miss (as is the case in every discipline that is lucky enough to be > based on journals) --- I would add that if we can move toward such a > system, it would probably be unnecessary to argue for the deadline > extensions that have such a detrimental effect on the work-life > balance of scientists (pace Talia's first point, which is well taken). > > I've missed submitting to POPL several times because the deadlines > didn't line up with the stage of my research, but I am dreaming of a > future where it really matters less for me that my research happens > to be "medium rare" on approximately July 1st each year. > > I also support and agree with everything that Gabriel has said in > this thread. I truly love workshops and seminars, and if I could > flush all these expensive and stressful conferences directly down > the toilet (together with the dogmatic ideology of our professional > organizations and their representatives, which we consume en gavage) > and instead just go to lovely low-stress workshops and send my work > periodically to journals, I would be so happy. > > Best, > Jon > > On Sun, Aug 23, 2020, at 11:11 AM, Talia Ringer wrote: >> [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >> >> I used to argue against changing conference deadline systems a lot, but the >> pandemic response and political events in the US have made it clear to me >> that this is a diversity issue. When a deadline is only once per year (and >> some of us do not have work that easily crosses over to other major >> conferences), missing it can be a major setback. And events like the >> pandemic have a disproportionate impact on groups that are traditionally >> underrepresented in our field. So the consequences of the deadline system >> are very uneven and reinforce our field's current demographic. >> >> I agree that it is absolutely prudent to take this opportunity to reflect >> on our review process. I want to push strongly for moving to a model in >> which deadlines exist but are much more frequent (say, monthly or >> quarterly). I think the approach Gabriel Scherer mentioned that is taken by >> "The Art, Science, and Engineering of Programming" would be better both for >> science and for diversity in our field. >> >> Talia >> https://dependenttyp.es/ >> >> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer >> wrote: >> >> [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list >> ] >> >> I am in broad agreement with many of Flavien's points. I hope that we can >> learn how to do virtual conferences well so that we can reduce our travel >> footprint in the future, not just due to pandemic issues. In this respect I >> have been fairly impressed with the degree of investment of many members of >> our community in finding and building better tools for virtual conferences. >> Thanks! >> >> I hope that this major change (that is imposed to us for an unpredictable >> amount of time) could also be an occasion to seriously consider >> de-synchronizing publication of our work from conference presentations. I >> think that journal publications have better academic review process, but >> we've been traditionally tied to major conferences as publication venues. >> Maybe it is time to change this? In this respect an interesting approach is >> "The Art, Science, and Engineering of Programming" journal which is coupled >> with the conference: journal publishes four volumes a year >> (trying to fit a three-months reviewing process), and the conference is >> held annually, with all papers accepted during the year presented. >> Forced-online venues could be an occasion to experiment with this. (We >> could think of other formats, such as having a *seminar* attached to a >> journal instead of a conference; so far I found it easier to enjoy online >> seminars than online conferences.) >> >> On Tue, Jul 28, 2020 at 2:47 PM Flavien Breuvart < >> breuvart at lipn.univ-paris13.fr> wrote: >> >>> [ The Types Forum, >> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>> ] >>> >>> Dear colleagues, >>> >>> This spring, under unfortunate circumstances, many conferences held >>> virtually. We have witnessed the disadvantages of such dispositives, but >>> also its numerous advantages. Many of those conferences have had >>> internal debates for debriefing this experiences, but I haven't seen any >>> large and public debate inside the community. I was hopping that some of >>> you may engage in such debates. >>> >>> As a starting point, I will try to succinctly expose my own point of >>> view, which is probably subjective, politically charged, and highly >>> debatable, but this is the whole point :-) >>> >>> I think we where all impressed by the high level of attendance of >>> conferences and workshops. But when thinking back at it, this situation >>> is perfectly normal as virtual conferences opened several blockades >>> usually preventing people from coming, in particular via the absence of >>> fees, the flexibility with respect to other duties (familial, teaching >>> or administrative), or the weight of travels. Even if this was the only >>> reason, I think it would be worth considering to secure part of these >>> improvements. >>> >>> Another, huge (but politically charged) advantage, is the drastic >>> reduction of the carbon footprint of our conferences. Several colleges >>> are advocating for a public engagement of the community to reduce its >>> global footprint. For example, see https://tcs4f.org/ which is a group >>> advocating for a 50% carbon reduction in theoretical computer sciences. >>> I have no doubt that other such initiative exist here and there; this >>> year unfortunate event at least showed that they are well founded and >>> not unreachable. >>> >>> That being said, I have to address the fact that our virtual conferences >>> had technical issues and that physical ones have several other >>> advantages. Concerning the technical issues (timeline clashes, internet >>> connection, organization...), I strongly believe that time and >>> experience can overcome most of them; I was helping in the organization >>> comity of FSCD and it appear that many issues could have been avoided by >>> a few technical adjustments (such as assigning two co-chairs for each >>> sessions for example). >>> >>> Concerning the advantages of conferences, I see three important ones : >>> 1) the chance encounters, 2) the strengthening of collaborations, and 3) >>> the prolonged focus. 1) From my (short) experience, the first can happen >>> in smaller scale meetings, that can be mostly local (with a minority of >>> invited non-local visitors). 2) The best way to strengthen >>> collaborations is not conferences but lab invitations (which could be >>> more frequent without conferences fees and time expenditures). 3) I got >>> the impression that most people where not as focus as in traditional >>> conferences, but not to a big margin, and mainly by lack of routine >>> (here I distinguish independent seminars and regular courses, as all >>> teachers I have seen the disaster of virtualization among our >> students...). >>> >>> All in all, I would advocate for more small scale meetings, more lab >>> invitations, but a virtualization of big scale conferences, and (why >>> not), the securing of some international virtual seminar that where very >>> interesting (thank you for the organizers that took those initiatives !). >>> >>> I hope I was not too long and too boring, do not hesitate to contradict >>> me, all I want is to start a fruitful debate. >>> >>> Best, >>> >>> Flavien >>> >>> >>> > > ? From oleg at okmij.org Mon Aug 24 05:59:55 2020 From: oleg at okmij.org (Oleg) Date: Mon, 24 Aug 2020 18:59:55 +0900 Subject: [TYPES] global debriefing over our virtual experience of conferences In-Reply-To: Message-ID: <20200824095955.GA3594@Melchior.localnet> Beside voicing my agreement with Gabriel and Jon and many others, I'd like to mention a point that I have not seen raised before. I have not seen any data as to the primary motivation for submitting to major conferences (and related symposia): -- submit to attend: people submit specifically to get a chance to attend a conference (I know that some people, especially students, won't get reimbursed if they don't present a paper). The people in this category would be content if there was a way to participate in formal and informal meetings without getting a paper accepted and without too much of a cost; -- submit to publish: people submit to get a paper *swiftly and competently* reviewed, and if successful, published in a well-read venue (the prestige of the venue is also not to be discounted: I've seen professors recommend to students, who are searching for a suitable topic, to look at the recent ICFP and POPL and PLDI. No journal and no `lesser' conference are recommended.) People in this category would be content if they didn't have to physically or virtually participate. I know the second category is not empty: it includes at least myself. What I (and probably other people in the same category, if exist) would wish is something like TOPLAS Letters. It doesn't have to be a shoddy venue: Physical Review Letters (which reviews and publishes rapidly and so imposes tight limits on the submission) is (one of) the most prestigious publications in Physics, probably more prestigious than Physical Review itself. I don't mean to supplant the workshops: just the other way around. A workshop would no longer need a big PC and (sometimes, too stressful) reviewing period: PC chairs merely select the appropriate papers already accepted for TOPLAS Letters and contact the authors if they are interested in presenting. At least the quality is already guaranteed, and so is the selection and the choice of topics. PC would spend more time then planning and promoting discussions and useful interactions. I have a more elaborate argumentation about the Letters. I'm not sure where to present it though: it is not actually on topic neither for this list nor the Climate list... Cheers, Oleg From monnier at iro.umontreal.ca Mon Aug 24 09:03:04 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Mon, 24 Aug 2020 09:03:04 -0400 Subject: [TYPES] online conferences should be free In-Reply-To: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> (Henning Basold's message of "Mon, 24 Aug 2020 08:48:59 +0200") References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> Message-ID: Henning Basold [2020-08-24 08:48:59] wrote: > I use the word "reproduce" purposefully because this kind of reality > descends merely from our society and is not an absolute truth. I believe the issue here is a bit different: it's the fact that if you pay less, you have less of an investment and that influences your behavior unconsciously. It does not just come from society but is much more ingrained in humans. I don't think we can realistically hope to avoid that effect on the premise that we're rational beings. BTW, the effect doesn't come only from money. It also comes for example from the investment (e.g. in time taken to organize the trip, ...) that comes with traveling to the conference. Stefan From h.basold at liacs.leidenuniv.nl Mon Aug 24 09:25:36 2020 From: h.basold at liacs.leidenuniv.nl (Henning Basold) Date: Mon, 24 Aug 2020 15:25:36 +0200 Subject: [TYPES] online conferences should be free In-Reply-To: References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> Message-ID: On 24/08/2020 15:03, Stefan Monnier wrote: > Henning Basold [2020-08-24 08:48:59] wrote: >> I use the word "reproduce" purposefully because this kind of reality >> descends merely from our society and is not an absolute truth. > > I believe the issue here is a bit different: it's the fact that if you > pay less, you have less of an investment and that influences your > behavior unconsciously. It does not just come from society but is much > more ingrained in humans. I don't think we can realistically hope to > avoid that effect on the premise that we're rational beings. > > BTW, the effect doesn't come only from money. It also comes for example > from the investment (e.g. in time taken to organize the trip, ...) that > comes with traveling to the conference. > > > Stefan > I agree with your analysis that any, as you call it, investment binds and influences us. All I objected to is that this investment has to come in form of monetary contributions and that this was presented as unavoidable reality. Instead, an investment can come, as you rightly say, in the form of time or community ties. The question is then if we need other incentives for participation, perhaps because conferences that need them can in fact be dispensed with, and what other incentives than money we could use in a world where travelling is not feasible on the same scale? Best, Henning From nicolai.kraus at gmail.com Mon Aug 24 12:18:17 2020 From: nicolai.kraus at gmail.com (Nicolai Kraus) Date: Mon, 24 Aug 2020 17:18:17 +0100 Subject: [TYPES] online conferences should be free In-Reply-To: References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> Message-ID: On Mon, Aug 24, 2020 at 2:25 PM Henning Basold wrote: > All I objected to is that this investment has to come > in form of monetary contributions and that this was presented as > unavoidable reality. Instead, an investment can come, as you rightly > say, in the form of time or community ties. > If one watches all pre-recorded talks (assuming something like at LICS - pre-recorded talks instead of live talks), one is more likely to participate actively. But this just shifts the problem, so I'm not sure if it helps in any way. At the same time, one can maybe from anonymous questionnaires or even simple view counts predict how well-prepared participants are and how seriously they will take the conference. I don't know whether that information could be used to improve the conference. Nicolai From sojakova.kristina at gmail.com Tue Aug 25 08:47:17 2020 From: sojakova.kristina at gmail.com (Kristina Sojakova) Date: Tue, 25 Aug 2020 14:47:17 +0200 Subject: [TYPES] online conferences should be free In-Reply-To: References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> Message-ID: <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> Dear all, Since Nicolai's post brought up pre-recorded talks:I found it extremely difficult to allocate enough time *prior* to the actual conference to watch almost any of the talks I was interested in - and there were about 10 I really wanted to see. Attending live talks (whether in person or online) during the conference days when the participation is the sole focus is very different from having to watch the talks beforehand; the latter essentially amounts to having a bunch of homework assigned. In the weeks prior to LICS I was relocating between continents and did not have time to watch any talks except one. This meant I was playing catch-up the entire time and was not able to get anything meaningful out of the conference itself. Hence I would most likely choose not to participate in this format again, regardless of how much it cost (or didn't). It is entirely possible however that I am in the minority and most participants were much better prepared. Best, Kristina On 8/24/2020 6:18 PM, Nicolai Kraus wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > On Mon, Aug 24, 2020 at 2:25 PM Henning Basold > wrote: > >> All I objected to is that this investment has to come >> in form of monetary contributions and that this was presented as >> unavoidable reality. Instead, an investment can come, as you rightly >> say, in the form of time or community ties. >> > If one watches all pre-recorded talks (assuming something like at LICS - > pre-recorded talks instead of live talks), one is more likely to > participate actively. But this just shifts the problem, so I'm not sure if > it helps in any way. At the same time, one can maybe from anonymous > questionnaires or even simple view counts predict how well-prepared > participants are and how seriously they will take the conference. I don't > know whether that information could be used to improve the conference. > Nicolai From gabriel.scherer at gmail.com Tue Aug 25 16:07:50 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Tue, 25 Aug 2020 22:07:50 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Thanks Mike for bringing clear arguments in favor. There was an interesting discussion with Nicolai, Henning and Stefan on the argument of registration fees helping engagement. I have some short points to make on this topic: 1. This exact argument (price signals value) is given by some high-ranked universities in the English-speaking part of the world that charge tens of thousands of dollars of tuition fee and *could* relatively easily cover the corresponding costs on their endowment money. 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 (for non-students) to help engagement by making the conference feel valuable, but now ICFP cost $100; one could argue for POPL'21 registration fees of $1000 for non-students, to make it *very* valuable as a conference. (As long as the people making pricing decisions are getting reimbursed for their registration fees, I guess we could pull this off? However, I think that this discussion on engagement is somewhat of a distraction. Having conference where people participate actively is certainly a good thing, it is *nice*. But sharing our knowledge and results in the most open way possible is *a core tenet of our duty as researchers*. For me it is clear that we should find *other* ways to help engagement that do not raise the barrier to entry to our conference, because the later is sensibly more important than the former. On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > Thanks for raising this issue. Just a few points about the other side of > the argument: > > It?s well known that things that are free are not valued (by humans) as > much as those that cost something, even a small amount. For example, see > Dan Ariely?s ?Predictably Irrational? which presents the results of several > experiments that demonstrate this. As a relevant case: Free MOOCs tended to > have lots of ?sign ups? but far fewer attendees, and even fewer completers. > > As such, if the goal is to have engaged attendees, trying to come closer > to the experience of traditional conferences, it might make sense to charge > something, even a small amount like $25, for at least some of the > population. This population might be people who have lots of social capital > already, and are generally busy, so they are more likely to blow off the > conference if they paid nothing for signing up. Such people might be those > that more junior attendees wish to meet. > > I note that engaged attendance was a goal when we had in-person > conferences, so I don?t see why we?d want to drop it now. Indeed, if people > don?t want to be engaged the videos will be available for free, afterward. > > Beyond the modest fees to run an online conference, which Talia mentions, > conference registration payments serve other purposes. Any surplus goes to > SIGPLAN, which turns around this surplus as good works, e.g., paying the > open access fees for PACMPL, which ICFP benefits from. It also makes > donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. > > Corporate sponsors can indeed pay some costs, but they also have > downsides. We are finding that many sponsors are not interested in > necessarily giving that much, and some are starting to make demands on how > the conference is run for their modest donation. This is a slippery slope > that the SIGPLAN EC is trying to avoid. > > Given that PLDI was completely free and ICFP followed a progressive fee > schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, > to see how the registration fee affected attendance. > > Thanks, > -Mike > > On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer > wrote: > >> [ The Types Forum, >> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >> >> >> >> I don't know about PLDI, but there are some costs associated with online >> >> events. For example, automatic captioning software is still not very good >> >> (Google's always turns "proofs" into "fruits" for me). Live captioning is >> >> really expensive! But it's also hugely important for disability >> >> accessibility. >> >> >> >> For students, ICFP was essentially free. I do agree that in principle, >> >> online conferences should be free, and online components of hybrid >> >> conferences should be free or strongly discounted. In practice, though, I >> >> do think that will mean finding sponsors for hidden costs that really are >> >> necessary. >> >> >> >> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >> gabriel.scherer at gmail.com> >> >> wrote: >> >> >> >> > [ The Types Forum, >> http://lists.seas.upenn.edu/mailman/listinfo/types-list >> >> > ] >> >> > >> >> > Dear types-list, >> >> > >> >> > Going on a tangent from Flavien's earlier post: I really think that >> online >> >> > conferences should be free. >> >> > >> >> > Several conferences (PLDI for example) managed to run free-of-charge >> since >> >> > the pandemic started, and they reported broader attendance and a strong >> >> > diversity of attendants, which sounds great. I don't think we can >> achieve >> >> > this with for-pay online conferences. >> >> > >> >> > ICFP is coming up shortly with a $100 registration price tag, and I did >> not >> >> > register. >> >> > >> >> > I'm aware that running a large virtual conference requires computing >> >> > resources that do have a cost. For PLDI for example, the report only >> says >> >> > that the cost was covered by industrial sponsors. Are numbers publicly >> >> > available on the cost of running a virtual conference? Note that if we >> >> > managed to run a conference on free software, I'm sure that institutions >> >> > and volunteers could be convinced to help hosting and monitoring the >> >> > conference services during the event. >> >> > >> >> From gabriel.scherer at gmail.com Tue Aug 25 16:24:51 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Tue, 25 Aug 2020 22:24:51 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > Beyond the modest fees to run an online conference, which Talia mentions, > conference registration payments serve other purposes. Any surplus goes to > SIGPLAN, which turns around this surplus as good works, e.g., paying the > open access fees for PACMPL, which ICFP benefits from. It also makes > donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. > On this point I have a simple request: could SIGPLAN release a budget summary for ICFP and PLDI this year? To have an informed discussion on online conference fees, we should know the amounts of the varied costs (including the "good works"), how much money comes in from sponsoring, and how much comes from conference fees. Since I raised the question of conference fees I have heard varied explanations from various people, for example the idea that PLDI was decided to have free registration before the actual conference-running costs were known, and that the $100 fee for ICFP adjusts to cover direct costs better. Your message rather suggests that the direct costs use only a modest fraction of the $100 (or maybe they can be covered entirely by sponsorship ?), but we cannot tell without any actual data. Others suggested that making the conference free was maybe doable for flagship conferences with an industrial presence such as ICFP or PLDI, but that this expectation could be problematic for more theoretical conferences with less sponsors. But then, FSCD and IJCAR reportedly ran fine without registration costs. (I don't think this is an outlandish or surprising request, for example I remember Adam Chlipala making exactly this request during a conference Town Hall a few years ago.) On Tue, Aug 25, 2020 at 10:07 PM Gabriel Scherer wrote: > Thanks Mike for bringing clear arguments in favor. > > There was an interesting discussion with Nicolai, Henning and Stefan on > the argument of registration fees helping engagement. I have some short > points to make on this topic: > > 1. This exact argument (price signals value) is given by some high-ranked > universities in the English-speaking part of the world that charge tens of > thousands of dollars of tuition fee and *could* relatively easily cover the > corresponding costs on their endowment money. > > 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 > (for non-students) to help engagement by making the conference feel > valuable, but now ICFP cost $100; one could argue for POPL'21 registration > fees of $1000 for non-students, to make it *very* valuable as a conference. > (As long as the people making pricing decisions are getting reimbursed for > their registration fees, I guess we could pull this off? > > However, I think that this discussion on engagement is somewhat of a > distraction. Having conference where people participate actively is > certainly a good thing, it is *nice*. But sharing our knowledge and results > in the most open way possible is *a core tenet of our duty as researchers*. > For me it is clear that we should find *other* ways to help engagement that > do not raise the barrier to entry to our conference, because the later is > sensibly more important than the former. > > > > > > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > >> Thanks for raising this issue. Just a few points about the other side of >> the argument: >> >> It?s well known that things that are free are not valued (by humans) as >> much as those that cost something, even a small amount. For example, see >> Dan Ariely?s ?Predictably Irrational? which presents the results of several >> experiments that demonstrate this. As a relevant case: Free MOOCs tended to >> have lots of ?sign ups? but far fewer attendees, and even fewer completers. >> >> As such, if the goal is to have engaged attendees, trying to come closer >> to the experience of traditional conferences, it might make sense to charge >> something, even a small amount like $25, for at least some of the >> population. This population might be people who have lots of social capital >> already, and are generally busy, so they are more likely to blow off the >> conference if they paid nothing for signing up. Such people might be those >> that more junior attendees wish to meet. >> >> I note that engaged attendance was a goal when we had in-person >> conferences, so I don?t see why we?d want to drop it now. Indeed, if people >> don?t want to be engaged the videos will be available for free, afterward. >> >> Beyond the modest fees to run an online conference, which Talia mentions, >> conference registration payments serve other purposes. Any surplus goes to >> SIGPLAN, which turns around this surplus as good works, e.g., paying the >> open access fees for PACMPL, which ICFP benefits from. It also makes >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >> >> Corporate sponsors can indeed pay some costs, but they also have >> downsides. We are finding that many sponsors are not interested in >> necessarily giving that much, and some are starting to make demands on how >> the conference is run for their modest donation. This is a slippery slope >> that the SIGPLAN EC is trying to avoid. >> >> Given that PLDI was completely free and ICFP followed a progressive fee >> schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, >> to see how the registration fee affected attendance. >> >> Thanks, >> -Mike >> >> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer >> wrote: >> >>> [ The Types Forum, >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >>> >>> >>> >>> I don't know about PLDI, but there are some costs associated with online >>> >>> events. For example, automatic captioning software is still not very good >>> >>> (Google's always turns "proofs" into "fruits" for me). Live captioning is >>> >>> really expensive! But it's also hugely important for disability >>> >>> accessibility. >>> >>> >>> >>> For students, ICFP was essentially free. I do agree that in principle, >>> >>> online conferences should be free, and online components of hybrid >>> >>> conferences should be free or strongly discounted. In practice, though, I >>> >>> do think that will mean finding sponsors for hidden costs that really are >>> >>> necessary. >>> >>> >>> >>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >>> gabriel.scherer at gmail.com> >>> >>> wrote: >>> >>> >>> >>> > [ The Types Forum, >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>> >>> > ] >>> >>> > >>> >>> > Dear types-list, >>> >>> > >>> >>> > Going on a tangent from Flavien's earlier post: I really think that >>> online >>> >>> > conferences should be free. >>> >>> > >>> >>> > Several conferences (PLDI for example) managed to run free-of-charge >>> since >>> >>> > the pandemic started, and they reported broader attendance and a strong >>> >>> > diversity of attendants, which sounds great. I don't think we can >>> achieve >>> >>> > this with for-pay online conferences. >>> >>> > >>> >>> > ICFP is coming up shortly with a $100 registration price tag, and I >>> did not >>> >>> > register. >>> >>> > >>> >>> > I'm aware that running a large virtual conference requires computing >>> >>> > resources that do have a cost. For PLDI for example, the report only >>> says >>> >>> > that the cost was covered by industrial sponsors. Are numbers publicly >>> >>> > available on the cost of running a virtual conference? Note that if we >>> >>> > managed to run a conference on free software, I'm sure that >>> institutions >>> >>> > and volunteers could be convinced to help hosting and monitoring the >>> >>> > conference services during the event. >>> >>> > >>> >>> From reuben.rowe at rhul.ac.uk Tue Aug 25 17:08:17 2020 From: reuben.rowe at rhul.ac.uk (Reuben Rowe) Date: Tue, 25 Aug 2020 22:08:17 +0100 Subject: [TYPES] [EXT] Re: online conferences should be free In-Reply-To: <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> Message-ID: <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> Dear Colleagues, I think this comment gets to the heart of one of the major issues with virtual conferences. I did think the LICS model of having talks pre-recorded and reserving synchronous sessions for questions and live chat worked extremely well to mitigate the very real fatigue of having to sit in front of one's screen watching hour upon hour of talks. However, as Kristina has pointed out (and this is my experience as well), it is often difficult to fully commit to being present, and also to "do the homework". In particular, I am at home, I have not booked time off specifically for the conference, I have on-going other responsibilities, my colleagues contact me about this and that, etc. The temptation is to assume that one can log on to the scheduled sessions, and get on with other stuff in the meantime. However, I think it is much harder to justify (perhaps both to one's employer and also to oneself) ignoring the non-conference commitments when there exists, at least ostensibly, the opportunity to fulfill them. One clear advantage of having to travel to a conference and attend in person is that one is automatically, a priori, "available" to engage. I'm not sure what can be done to mitigate this. Perhaps it is just the latent cost of the virtual model, and if we want to attend a virtual conference we must just take responsibility to ensure that we engage. Reuben On 25/08/2020 13:47, Kristina Sojakova wrote: > [ The Types Forum, > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 > ] > > Dear all, > > Since Nicolai's post brought up pre-recorded talks:I found it > extremely difficult to allocate enough time *prior* to the actual > conference to watch almost any of the talks I was interested in - and > there were about 10 I really wanted to see. Attending live talks > (whether in person or online) during the conference days when the > participation is the sole focus is very different from having to watch > the talks beforehand; the latter essentially amounts to having a bunch > of homework assigned. > > In the weeks prior to LICS I was relocating between continents and did > not have time to watch any talks except one. This meant I was playing > catch-up the entire time and was not able to get anything meaningful > out of the conference itself. Hence I would most likely choose not to > participate in this format again, regardless of how much it cost (or > didn't). It is entirely possible however that I am in the minority and > most participants were much better prepared. > > Best, > > Kristina > > > > On 8/24/2020 6:18 PM, Nicolai Kraus wrote: >> [ The Types Forum, >> https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 >> ] >> >> On Mon, Aug 24, 2020 at 2:25 PM Henning Basold >> >> wrote: >> >>> All I objected to is that this investment has to come >>> in form of monetary contributions and that this was presented as >>> unavoidable reality. Instead, an investment can come, as you rightly >>> say, in the form of time or community ties. >>> >> If one watches all pre-recorded talks (assuming something like at LICS - >> pre-recorded talks instead of live talks), one is more likely to >> participate actively. But this just shifts the problem, so I'm not >> sure if >> it helps in any way. At the same time, one can maybe from anonymous >> questionnaires or even simple view counts predict how well-prepared >> participants are and how seriously they will take the conference. I >> don't >> know whether that information could be used to improve the conference. >> Nicolai This email, its contents and any attachments are intended solely for the addressee and may contain confidential information. In certain circumstances, it may also be subject to legal privilege. Any unauthorised use, disclosure, or copying is not permitted. If you have received this email in error, please notify us and immediately and permanently delete it. Any views or opinions expressed in personal emails are solely those of the author and do not necessarily represent those of Royal Holloway, University of London. It is your responsibility to ensure that this email and any attachments are virus free. From monnier at iro.umontreal.ca Tue Aug 25 18:37:32 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Tue, 25 Aug 2020 18:37:32 -0400 Subject: [TYPES] online conferences should be free In-Reply-To: (Gabriel Scherer's message of "Tue, 25 Aug 2020 22:07:50 +0200") References: Message-ID: Gabriel Scherer [2020-08-25 22:07:50] wrote: > However, I think that this discussion on engagement is somewhat of a > distraction. Having conference where people participate actively is > certainly a good thing, it is *nice*. But sharing our knowledge and results > in the most open way possible is *a core tenet of our duty as researchers*. Partly, yes, but then journals would be all we need and we could ditch conferences, right? Despite appearances I'm not actually arguing against making conferences free, it's just that we need to take into account some of the effects on us imperfect humans. BTW, while watching ICFP, somewhat pleased with Clowdr [ for a first run of the software, I'm really pleased ] but annoyed at some aspects [ besides the need to run proprietary code for Zoom ] such as the fact that I can't find a recording of the TyDe talks I missed... ..."I had a dream" of an ACM library where each paper would link not just to the PDF but also to the corresponding talk(s?) [ which we could watch without having to get Google involved because it would be hosted on, say, a Peertube instance run by the ACM ], and also to a recording of the Q&A session(s), as well as some way to post questions/answers/comments/annotations, moderated by ourselves (along a model similar to Stackexchange, maybe). In this dream, I'm not sure what conferences would look like, but maybe they would tie the live Q&A sessions less tightly to the talks (so you can watch the talks that entice you at your own rhythm but still get to ask questions). Stefan From julbinb at gmail.com Tue Aug 25 19:12:17 2020 From: julbinb at gmail.com (Julia Belyakova) Date: Tue, 25 Aug 2020 19:12:17 -0400 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: To add on the topic of value for money. On one hand, I agree that people who pay might participate more actively and consume more of the conference content. On the other hand, if a lot more people join a conference for free but participate in a small number of events, that does not seem to be a necessarily bad thing. Of course, if 1000 people register and nobody participates in anything, that's a disaster. But if out of those 1000, you have 100 active people at a time, it's probably no worse (or even better) than 50 fully committed participants at all times. -- Kind regards, Julia ??, 25 ???. 2020 ?. ? 16:27, Gabriel Scherer : > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > > > Beyond the modest fees to run an online conference, which Talia mentions, > > conference registration payments serve other purposes. Any surplus goes > to > > SIGPLAN, which turns around this surplus as good works, e.g., paying the > > open access fees for PACMPL, which ICFP benefits from. It also makes > > donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. > > > > On this point I have a simple request: could SIGPLAN release a budget > summary for ICFP and PLDI this year? To have an informed discussion on > online conference fees, we should know the amounts of the varied costs > (including the "good works"), how much money comes in from sponsoring, and > how much comes from conference fees. > > Since I raised the question of conference fees I have heard varied > explanations from various people, for example the idea that PLDI was > decided to have free registration before the actual conference-running > costs were known, and that the $100 fee for ICFP adjusts to cover direct > costs better. Your message rather suggests that the direct costs use only a > modest fraction of the $100 (or maybe they can be covered entirely by > sponsorship ?), but we cannot tell without any actual data. Others > suggested that making the conference free was maybe doable for flagship > conferences with an industrial presence such as ICFP or PLDI, but that this > expectation could be problematic for more theoretical conferences with less > sponsors. But then, FSCD and IJCAR reportedly ran fine without registration > costs. > > (I don't think this is an outlandish or surprising request, for example I > remember Adam Chlipala making exactly this request during a conference Town > Hall a few years ago.) > > On Tue, Aug 25, 2020 at 10:07 PM Gabriel Scherer < > gabriel.scherer at gmail.com> > wrote: > > > Thanks Mike for bringing clear arguments in favor. > > > > There was an interesting discussion with Nicolai, Henning and Stefan on > > the argument of registration fees helping engagement. I have some short > > points to make on this topic: > > > > 1. This exact argument (price signals value) is given by some high-ranked > > universities in the English-speaking part of the world that charge tens > of > > thousands of dollars of tuition fee and *could* relatively easily cover > the > > corresponding costs on their endowment money. > > > > 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 > > (for non-students) to help engagement by making the conference feel > > valuable, but now ICFP cost $100; one could argue for POPL'21 > registration > > fees of $1000 for non-students, to make it *very* valuable as a > conference. > > (As long as the people making pricing decisions are getting reimbursed > for > > their registration fees, I guess we could pull this off? > > > > However, I think that this discussion on engagement is somewhat of a > > distraction. Having conference where people participate actively is > > certainly a good thing, it is *nice*. But sharing our knowledge and > results > > in the most open way possible is *a core tenet of our duty as > researchers*. > > For me it is clear that we should find *other* ways to help engagement > that > > do not raise the barrier to entry to our conference, because the later is > > sensibly more important than the former. > > > > > > > > > > > > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > > > >> Thanks for raising this issue. Just a few points about the other side of > >> the argument: > >> > >> It?s well known that things that are free are not valued (by humans) as > >> much as those that cost something, even a small amount. For example, see > >> Dan Ariely?s ?Predictably Irrational? which presents the results of > several > >> experiments that demonstrate this. As a relevant case: Free MOOCs > tended to > >> have lots of ?sign ups? but far fewer attendees, and even fewer > completers. > >> > >> As such, if the goal is to have engaged attendees, trying to come closer > >> to the experience of traditional conferences, it might make sense to > charge > >> something, even a small amount like $25, for at least some of the > >> population. This population might be people who have lots of social > capital > >> already, and are generally busy, so they are more likely to blow off the > >> conference if they paid nothing for signing up. Such people might be > those > >> that more junior attendees wish to meet. > >> > >> I note that engaged attendance was a goal when we had in-person > >> conferences, so I don?t see why we?d want to drop it now. Indeed, if > people > >> don?t want to be engaged the videos will be available for free, > afterward. > >> > >> Beyond the modest fees to run an online conference, which Talia > mentions, > >> conference registration payments serve other purposes. Any surplus goes > to > >> SIGPLAN, which turns around this surplus as good works, e.g., paying the > >> open access fees for PACMPL, which ICFP benefits from. It also makes > >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. > >> > >> Corporate sponsors can indeed pay some costs, but they also have > >> downsides. We are finding that many sponsors are not interested in > >> necessarily giving that much, and some are starting to make demands on > how > >> the conference is run for their modest donation. This is a slippery > slope > >> that the SIGPLAN EC is trying to avoid. > >> > >> Given that PLDI was completely free and ICFP followed a progressive fee > >> schedule, I?ll be curious to compare the ICFP outbrief with that of > PLDI?s, > >> to see how the registration fee affected attendance. > >> > >> Thanks, > >> -Mike > >> > >> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer < > tringer at cs.washington.edu> > >> wrote: > >> > >>> [ The Types Forum, > >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > >>> > >>> > >>> > >>> I don't know about PLDI, but there are some costs associated with > online > >>> > >>> events. For example, automatic captioning software is still not very > good > >>> > >>> (Google's always turns "proofs" into "fruits" for me). Live captioning > is > >>> > >>> really expensive! But it's also hugely important for disability > >>> > >>> accessibility. > >>> > >>> > >>> > >>> For students, ICFP was essentially free. I do agree that in principle, > >>> > >>> online conferences should be free, and online components of hybrid > >>> > >>> conferences should be free or strongly discounted. In practice, > though, I > >>> > >>> do think that will mean finding sponsors for hidden costs that really > are > >>> > >>> necessary. > >>> > >>> > >>> > >>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < > >>> gabriel.scherer at gmail.com> > >>> > >>> wrote: > >>> > >>> > >>> > >>> > [ The Types Forum, > >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list > >>> > >>> > ] > >>> > >>> > > >>> > >>> > Dear types-list, > >>> > >>> > > >>> > >>> > Going on a tangent from Flavien's earlier post: I really think that > >>> online > >>> > >>> > conferences should be free. > >>> > >>> > > >>> > >>> > Several conferences (PLDI for example) managed to run free-of-charge > >>> since > >>> > >>> > the pandemic started, and they reported broader attendance and a > strong > >>> > >>> > diversity of attendants, which sounds great. I don't think we can > >>> achieve > >>> > >>> > this with for-pay online conferences. > >>> > >>> > > >>> > >>> > ICFP is coming up shortly with a $100 registration price tag, and I > >>> did not > >>> > >>> > register. > >>> > >>> > > >>> > >>> > I'm aware that running a large virtual conference requires computing > >>> > >>> > resources that do have a cost. For PLDI for example, the report only > >>> says > >>> > >>> > that the cost was covered by industrial sponsors. Are numbers > publicly > >>> > >>> > available on the cost of running a virtual conference? Note that if > we > >>> > >>> > managed to run a conference on free software, I'm sure that > >>> institutions > >>> > >>> > and volunteers could be convinced to help hosting and monitoring the > >>> > >>> > conference services during the event. > >>> > >>> > > >>> > >>> > From nicolai.kraus at gmail.com Wed Aug 26 03:21:46 2020 From: nicolai.kraus at gmail.com (Nicolai Kraus) Date: Wed, 26 Aug 2020 08:21:46 +0100 Subject: [TYPES] [EXT] Re: online conferences should be free In-Reply-To: <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> Message-ID: On Wed, Aug 26, 2020 at 7:02 AM Reuben Rowe wrote: > However, as Kristina has pointed out (and this is my experience as > well), it is often difficult to fully commit to being present, and also > to "do the homework". In particular, I am at home, I have not booked > time off specifically for the conference, I have on-going other > responsibilities, my colleagues contact me about this and that, etc. > I agree that this is true at the moment, but I don't see why it has to stay true in the future. Why can we not book time off for an online conference? Why do we feel free from other on-going commitments when we are at a physical conference but not when we are at an online conference? Why can we not "do the homework" when we expect our students to do ti? Whenever things change, we have to adapt and change our attitudes. Maybe that's hard, but I don't think it is as difficult as solving the issue of global warming. There are many small and easily implementable things that could help us. For example, some universities have "writing retreats" where one can go to some other place (which can actually be very close) and write without being distracted; I have never tried it, but I am very happy to believe that it works. Similarly, a university could offer a bookable special room for online conferences, so one would be "somewhere else," and the rule would be that whenever one has booked the room, one should be treated as if one was at a physical conference. If finding the time for watching talks is an issue, one could simply let the virtual conference start two days earlier and keep the first two days completely free so that everyone who hasn't done it at that point can catch up with the talks. I know that everything you said is a real issue (which I also know from my own experience), but I don't believe that these things are unsolvable problems. > The temptation is to assume that one can log on to the scheduled > sessions, and get on with other stuff in the meantime. However, I think > it is much harder to justify (perhaps both to one's employer and also to > oneself) ignoring the non-conference commitments when there exists, at > least ostensibly, the opportunity to fulfill them. > Yes, of course I know what you mean; but this "only" requires a general attitude change, which I think is definitely possible. Nicolai > > One clear advantage of having to travel to a conference and attend in > person is that one is automatically, a priori, "available" to engage. > > I'm not sure what can be done to mitigate this. Perhaps it is just the > latent cost of the virtual model, and if we want to attend a virtual > conference we must just take responsibility to ensure that we engage. > > Reuben > > On 25/08/2020 13:47, Kristina Sojakova wrote: > > [ The Types Forum, > > > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 > > ] > > > > Dear all, > > > > Since Nicolai's post brought up pre-recorded talks:I found it > > extremely difficult to allocate enough time *prior* to the actual > > conference to watch almost any of the talks I was interested in - and > > there were about 10 I really wanted to see. Attending live talks > > (whether in person or online) during the conference days when the > > participation is the sole focus is very different from having to watch > > the talks beforehand; the latter essentially amounts to having a bunch > > of homework assigned. > > > > In the weeks prior to LICS I was relocating between continents and did > > not have time to watch any talks except one. This meant I was playing > > catch-up the entire time and was not able to get anything meaningful > > out of the conference itself. Hence I would most likely choose not to > > participate in this format again, regardless of how much it cost (or > > didn't). It is entirely possible however that I am in the minority and > > most participants were much better prepared. > > > > Best, > > > > Kristina > > > > > > > > On 8/24/2020 6:18 PM, Nicolai Kraus wrote: > >> [ The Types Forum, > >> > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 > >> ] > >> > >> On Mon, Aug 24, 2020 at 2:25 PM Henning Basold > >> > >> wrote: > >> > >>> All I objected to is that this investment has to come > >>> in form of monetary contributions and that this was presented as > >>> unavoidable reality. Instead, an investment can come, as you rightly > >>> say, in the form of time or community ties. > >>> > >> If one watches all pre-recorded talks (assuming something like at LICS - > >> pre-recorded talks instead of live talks), one is more likely to > >> participate actively. But this just shifts the problem, so I'm not > >> sure if > >> it helps in any way. At the same time, one can maybe from anonymous > >> questionnaires or even simple view counts predict how well-prepared > >> participants are and how seriously they will take the conference. I > >> don't > >> know whether that information could be used to improve the conference. > >> Nicolai > This email, its contents and any attachments are intended solely for the > addressee and may contain confidential information. In certain > circumstances, it may also be subject to legal privilege. Any unauthorised > use, disclosure, or copying is not permitted. If you have received this > email in error, please notify us and immediately and permanently delete it. > Any views or opinions expressed in personal emails are solely those of the > author and do not necessarily represent those of Royal Holloway, University > of London. It is your responsibility to ensure that this email and any > attachments are virus free. > From harley.eades at gmail.com Wed Aug 26 10:55:12 2020 From: harley.eades at gmail.com (Harley D. Eades III) Date: Wed, 26 Aug 2020 10:55:12 -0400 Subject: [TYPES] [EXT] Re: online conferences should be free In-Reply-To: <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> Message-ID: Hi, all. I would like to also add to this line of thought. As someone with ADHD I find virtual conferences, esp. in the model of ICFP where there is a full day of being on a computer, hard to handle without getting distracted. The fact that the talks are recorded means I can get caught up on them, but the essential part of engaging with others during the conference has been difficult for me. This also adds in the anxiety of it all; the lack in engagement means the lack of being seen in the community which can have a negative impact. When I go to a conference, I plan my program, and then I'm there, my brain knows I'm there, so it engages a lot easier, and the stress goes down. I think spreading virtual conferences out might help with a lot of this. Nothing says we have to have conferences for full days for a week. We could block a day of the week for a month. For example, every Wed. is LICS or ICFP day, and then this would help mitigate the problems we are talking about. One comment about record talks. I myself love recorded talks, but only for the ones I know are relevant to what I'm doing. So I make time to watch those videos, but I'm far less likely to watch a talk I feel might not be of interest to me. This is not a good thing, because I could miss out on something interesting, or new connections to my own work etc. But, when at the conference it is far easier to watch these talks, because I go to the sessions in person, and I tend to not leave a session early. Furthermore, I can say that employers, in my experience, do not consider a virtual conference in the same way as an in-person one. I am still teaching, mentoring, holding office hours, etc. This takes away from my ability to engage. This is a great discussion, I've learned a lot from it! Best, Harley On Wed, Aug 26, 2020 at 2:02 AM Reuben Rowe wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list > ] > > Dear Colleagues, > > I think this comment gets to the heart of one of the major issues with > virtual conferences. > > I did think the LICS model of having talks pre-recorded and reserving > synchronous sessions for questions and live chat worked extremely well > to mitigate the very real fatigue of having to sit in front of one's > screen watching hour upon hour of talks. > > However, as Kristina has pointed out (and this is my experience as > well), it is often difficult to fully commit to being present, and also > to "do the homework". In particular, I am at home, I have not booked > time off specifically for the conference, I have on-going other > responsibilities, my colleagues contact me about this and that, etc. > > The temptation is to assume that one can log on to the scheduled > sessions, and get on with other stuff in the meantime. However, I think > it is much harder to justify (perhaps both to one's employer and also to > oneself) ignoring the non-conference commitments when there exists, at > least ostensibly, the opportunity to fulfill them. > > One clear advantage of having to travel to a conference and attend in > person is that one is automatically, a priori, "available" to engage. > > I'm not sure what can be done to mitigate this. Perhaps it is just the > latent cost of the virtual model, and if we want to attend a virtual > conference we must just take responsibility to ensure that we engage. > > Reuben > > On 25/08/2020 13:47, Kristina Sojakova wrote: > > [ The Types Forum, > > > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 > > ] > > > > Dear all, > > > > Since Nicolai's post brought up pre-recorded talks:I found it > > extremely difficult to allocate enough time *prior* to the actual > > conference to watch almost any of the talks I was interested in - and > > there were about 10 I really wanted to see. Attending live talks > > (whether in person or online) during the conference days when the > > participation is the sole focus is very different from having to watch > > the talks beforehand; the latter essentially amounts to having a bunch > > of homework assigned. > > > > In the weeks prior to LICS I was relocating between continents and did > > not have time to watch any talks except one. This meant I was playing > > catch-up the entire time and was not able to get anything meaningful > > out of the conference itself. Hence I would most likely choose not to > > participate in this format again, regardless of how much it cost (or > > didn't). It is entirely possible however that I am in the minority and > > most participants were much better prepared. > > > > Best, > > > > Kristina > > > > > > > > On 8/24/2020 6:18 PM, Nicolai Kraus wrote: > >> [ The Types Forum, > >> > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=02%7C01%7Creuben.rowe%40rhul.ac.uk%7C82c5414d38a845c4736f08d8492f927a%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637339815963836441&sdata=x30U4UNN1cRwRhsSo9XU42QlHvq1INlK8K%2F%2BhGb48Yo%3D&reserved=0 > >> ] > >> > >> On Mon, Aug 24, 2020 at 2:25 PM Henning Basold > >> > >> wrote: > >> > >>> All I objected to is that this investment has to come > >>> in form of monetary contributions and that this was presented as > >>> unavoidable reality. Instead, an investment can come, as you rightly > >>> say, in the form of time or community ties. > >>> > >> If one watches all pre-recorded talks (assuming something like at LICS - > >> pre-recorded talks instead of live talks), one is more likely to > >> participate actively. But this just shifts the problem, so I'm not > >> sure if > >> it helps in any way. At the same time, one can maybe from anonymous > >> questionnaires or even simple view counts predict how well-prepared > >> participants are and how seriously they will take the conference. I > >> don't > >> know whether that information could be used to improve the conference. > >> Nicolai > This email, its contents and any attachments are intended solely for the > addressee and may contain confidential information. In certain > circumstances, it may also be subject to legal privilege. Any unauthorised > use, disclosure, or copying is not permitted. If you have received this > email in error, please notify us and immediately and permanently delete it. > Any views or opinions expressed in personal emails are solely those of the > author and do not necessarily represent those of Royal Holloway, University > of London. It is your responsibility to ensure that this email and any > attachments are virus free. > From hendrik at topoi.pooq.com Wed Aug 26 15:09:26 2020 From: hendrik at topoi.pooq.com (Hendrik Boom) Date: Wed, 26 Aug 2020 15:09:26 -0400 Subject: [TYPES] Online vs travel In-Reply-To: <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> References: <59dd6435-591a-503a-870b-ea2968594b30@liacs.leidenuniv.nl> <703adcf4-335c-c45e-f9a6-f75e5934f0b4@gmail.com> <2b0b345e-479c-7e22-2d70-69f96910897c@rhul.ac.uk> Message-ID: <20200826190925.x3d5p6wsavxoqln2@topoi.pooq.com> On Tue, Aug 25, 2020 at 10:08:17PM +0100, Reuben Rowe wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Dear Colleagues, > > I think this comment gets to the heart of one of the major issues with > virtual conferences. > > I did think the LICS model of having talks pre-recorded and reserving > synchronous sessions for questions and live chat worked extremely well > to mitigate the very real fatigue of having to sit in front of one's > screen watching hour upon hour of talks. Indeed. I don't manage to pay full attention like that. I tire, and I cannot fully appreciate the talks. In the days I still had a travel budget, The same would happen in a physical-presence meeting. Not to mention effectively missing half the talks altogether in the first few days because of jet lag. (jet lag isn't as bad if nighttime doesn't shift to a new schedule) A big advantage in an online conference is that I do not end up attending all the talks. I do not become exhausted attending talks of no interest to me. And I have attention left over for the talks that are of interest. And if too much is of interest (I should be so lucky!) I can *still* attend the talk later via the recording. Nowadays, I have essentially no travel budget. Most of the events I now attend online I would be completely unable to attend if I had to be physically present. -- hendrik From jung at mpi-sws.org Thu Aug 27 04:08:55 2020 From: jung at mpi-sws.org (Ralf Jung) Date: Thu, 27 Aug 2020 10:08:55 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Hi all, on the topic of conference fees, I was just made aware of this interesting blog post: https://andreas-zeller.info/2018/02/01/where-your-conference-fees-go-to.html It's pre-Covid, and the conference is smaller than POPL/ICFP/PLDI, but this could still provide some useful data. Kind regards, Ralf On 25.08.20 22:24, Gabriel Scherer wrote: > [ The Types Forum, http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > >> Beyond the modest fees to run an online conference, which Talia mentions, >> conference registration payments serve other purposes. Any surplus goes to >> SIGPLAN, which turns around this surplus as good works, e.g., paying the >> open access fees for PACMPL, which ICFP benefits from. It also makes >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >> > > On this point I have a simple request: could SIGPLAN release a budget > summary for ICFP and PLDI this year? To have an informed discussion on > online conference fees, we should know the amounts of the varied costs > (including the "good works"), how much money comes in from sponsoring, and > how much comes from conference fees. > > Since I raised the question of conference fees I have heard varied > explanations from various people, for example the idea that PLDI was > decided to have free registration before the actual conference-running > costs were known, and that the $100 fee for ICFP adjusts to cover direct > costs better. Your message rather suggests that the direct costs use only a > modest fraction of the $100 (or maybe they can be covered entirely by > sponsorship ?), but we cannot tell without any actual data. Others > suggested that making the conference free was maybe doable for flagship > conferences with an industrial presence such as ICFP or PLDI, but that this > expectation could be problematic for more theoretical conferences with less > sponsors. But then, FSCD and IJCAR reportedly ran fine without registration > costs. > > (I don't think this is an outlandish or surprising request, for example I > remember Adam Chlipala making exactly this request during a conference Town > Hall a few years ago.) > > On Tue, Aug 25, 2020 at 10:07 PM Gabriel Scherer > wrote: > >> Thanks Mike for bringing clear arguments in favor. >> >> There was an interesting discussion with Nicolai, Henning and Stefan on >> the argument of registration fees helping engagement. I have some short >> points to make on this topic: >> >> 1. This exact argument (price signals value) is given by some high-ranked >> universities in the English-speaking part of the world that charge tens of >> thousands of dollars of tuition fee and *could* relatively easily cover the >> corresponding costs on their endowment money. >> >> 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 >> (for non-students) to help engagement by making the conference feel >> valuable, but now ICFP cost $100; one could argue for POPL'21 registration >> fees of $1000 for non-students, to make it *very* valuable as a conference. >> (As long as the people making pricing decisions are getting reimbursed for >> their registration fees, I guess we could pull this off? >> >> However, I think that this discussion on engagement is somewhat of a >> distraction. Having conference where people participate actively is >> certainly a good thing, it is *nice*. But sharing our knowledge and results >> in the most open way possible is *a core tenet of our duty as researchers*. >> For me it is clear that we should find *other* ways to help engagement that >> do not raise the barrier to entry to our conference, because the later is >> sensibly more important than the former. >> >> >> >> >> >> On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: >> >>> Thanks for raising this issue. Just a few points about the other side of >>> the argument: >>> >>> It?s well known that things that are free are not valued (by humans) as >>> much as those that cost something, even a small amount. For example, see >>> Dan Ariely?s ?Predictably Irrational? which presents the results of several >>> experiments that demonstrate this. As a relevant case: Free MOOCs tended to >>> have lots of ?sign ups? but far fewer attendees, and even fewer completers. >>> >>> As such, if the goal is to have engaged attendees, trying to come closer >>> to the experience of traditional conferences, it might make sense to charge >>> something, even a small amount like $25, for at least some of the >>> population. This population might be people who have lots of social capital >>> already, and are generally busy, so they are more likely to blow off the >>> conference if they paid nothing for signing up. Such people might be those >>> that more junior attendees wish to meet. >>> >>> I note that engaged attendance was a goal when we had in-person >>> conferences, so I don?t see why we?d want to drop it now. Indeed, if people >>> don?t want to be engaged the videos will be available for free, afterward. >>> >>> Beyond the modest fees to run an online conference, which Talia mentions, >>> conference registration payments serve other purposes. Any surplus goes to >>> SIGPLAN, which turns around this surplus as good works, e.g., paying the >>> open access fees for PACMPL, which ICFP benefits from. It also makes >>> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >>> >>> Corporate sponsors can indeed pay some costs, but they also have >>> downsides. We are finding that many sponsors are not interested in >>> necessarily giving that much, and some are starting to make demands on how >>> the conference is run for their modest donation. This is a slippery slope >>> that the SIGPLAN EC is trying to avoid. >>> >>> Given that PLDI was completely free and ICFP followed a progressive fee >>> schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, >>> to see how the registration fee affected attendance. >>> >>> Thanks, >>> -Mike >>> >>> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer >>> wrote: >>> >>>> [ The Types Forum, >>>> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >>>> >>>> >>>> >>>> I don't know about PLDI, but there are some costs associated with online >>>> >>>> events. For example, automatic captioning software is still not very good >>>> >>>> (Google's always turns "proofs" into "fruits" for me). Live captioning is >>>> >>>> really expensive! But it's also hugely important for disability >>>> >>>> accessibility. >>>> >>>> >>>> >>>> For students, ICFP was essentially free. I do agree that in principle, >>>> >>>> online conferences should be free, and online components of hybrid >>>> >>>> conferences should be free or strongly discounted. In practice, though, I >>>> >>>> do think that will mean finding sponsors for hidden costs that really are >>>> >>>> necessary. >>>> >>>> >>>> >>>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >>>> gabriel.scherer at gmail.com> >>>> >>>> wrote: >>>> >>>> >>>> >>>>> [ The Types Forum, >>>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>>> >>>>> ] >>>> >>>>> >>>> >>>>> Dear types-list, >>>> >>>>> >>>> >>>>> Going on a tangent from Flavien's earlier post: I really think that >>>> online >>>> >>>>> conferences should be free. >>>> >>>>> >>>> >>>>> Several conferences (PLDI for example) managed to run free-of-charge >>>> since >>>> >>>>> the pandemic started, and they reported broader attendance and a strong >>>> >>>>> diversity of attendants, which sounds great. I don't think we can >>>> achieve >>>> >>>>> this with for-pay online conferences. >>>> >>>>> >>>> >>>>> ICFP is coming up shortly with a $100 registration price tag, and I >>>> did not >>>> >>>>> register. >>>> >>>>> >>>> >>>>> I'm aware that running a large virtual conference requires computing >>>> >>>>> resources that do have a cost. For PLDI for example, the report only >>>> says >>>> >>>>> that the cost was covered by industrial sponsors. Are numbers publicly >>>> >>>>> available on the cost of running a virtual conference? Note that if we >>>> >>>>> managed to run a conference on free software, I'm sure that >>>> institutions >>>> >>>>> and volunteers could be convinced to help hosting and monitoring the >>>> >>>>> conference services during the event. >>>> >>>>> >>>> >>>> -- Website: https://people.mpi-sws.org/~jung/ From gabriel.scherer at gmail.com Tue Sep 8 07:04:43 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Tue, 8 Sep 2020 13:04:43 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Dear list, I would like to come back to a reasonable argument for non-free conference that was presented here and outside the list: there is a natural idea that even online conferences require servers, bandwidth, etc., which cost money, and that participants would pay this cost. However I would contend that (1) (despite the lack of numbers given by conference organizers) we should assume that those technical costs are fairly small, (2) the infrastructure to run online conferences could be made part of common services whose costs are mutualized, and (3) sponsors *and* our research institutions could cover those costs. (1): For conferences in our community that were forced to move online, being free is the norm, not the exception. PLDI was free, but so were LICS, FSCD, IJCAR, etc. This suggests that running free is financially possible/reasonable, and thus that the costs are not very high. (In addition to the technical cost of the hardware, electricity, network etc., there are of course human costs associated to having people maintain the service during the conference period, developing the services beforehand, and otherwise setup and run the conference. Those are probably higher than the technical costs (many aspects of it are done by academics for free, that is, mutualized by their employers). It's hard to say more without some transparency on online conference budgets. Then again, most conferences were able to run free.) (2): Suppose you go to a seminar talk at some university close by; before the talk, the organizer asks you to pay $20 for the upkeep of the seminar room, or leave -- you get a link to watch the recorded talk for free. Many of us would find the demand disturbing; in any case, while maintaining university buildings does come with large costs, we don't suggest that our colleagues coming to attend a seminar talk should pay for it. It is not obvious that attendants of an online research conference should be the ones paying the cost; it is one choice among several options, and I think the wrong choice. Another example: ICFP had 1100 registrants being nice to each other online for a week, but the #haskell IRC channel on Freenode has 1000+ users talking to each other all year long. We don't force them to pay for the Freenode servers before they can access the channel. (Or this mailing-list, whose costs I assume are generously covered by UPenn. We are surrounded by services provided by our community with mutualized costs, which make them much more vibrant, valuable, effective than if we asked their users to pay to use them.) (3): Personally I have mixed feelings about the comments on the dangers of relying too much on our sponsors. I certainly agree with the idea. But I'm surprised that it comes up now, that we have a chance to turn a fairly unpleasant conference situation into a least a big jump in accessibility of our research community (among other nice benefits), while I didn't hear very much of it in previous years, when conferences had much larger expenses that required sponsoring. ICFP rented an entire *museum* in Oxford for one evening, and then again in Berlin. Were people worrying about depending on sponsors then? Maybe they were and we just didn't have public discussions about it; but I think this concern is much less relevant in the context of online conferences whose costs are substantially lower than for a physical conference. If we had some transparent information on the budget required to run our online conferences, we could also go to our institutions to ask them to help. My employer, INRIA, has explicit procedures in place to support scientific events that require funding, I'm sure many others have (see the list of institutions supporting arxiv.org at https://arxiv.org/about/ourmembers ). The amount you typically get is lower than most offers from industrial sponsors; but many universities and research places around the world would be happy to participate in hosting and running our conferences, especially if it came with the explicit goal of making them much more widely accessible. Best On Tue, Aug 25, 2020 at 10:07 PM Gabriel Scherer wrote: > Thanks Mike for bringing clear arguments in favor. > > There was an interesting discussion with Nicolai, Henning and Stefan on > the argument of registration fees helping engagement. I have some short > points to make on this topic: > > 1. This exact argument (price signals value) is given by some high-ranked > universities in the English-speaking part of the world that charge tens of > thousands of dollars of tuition fee and *could* relatively easily cover the > corresponding costs on their endowment money. > > 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 > (for non-students) to help engagement by making the conference feel > valuable, but now ICFP cost $100; one could argue for POPL'21 registration > fees of $1000 for non-students, to make it *very* valuable as a conference. > (As long as the people making pricing decisions are getting reimbursed for > their registration fees, I guess we could pull this off? > > However, I think that this discussion on engagement is somewhat of a > distraction. Having conference where people participate actively is > certainly a good thing, it is *nice*. But sharing our knowledge and results > in the most open way possible is *a core tenet of our duty as researchers*. > For me it is clear that we should find *other* ways to help engagement that > do not raise the barrier to entry to our conference, because the later is > sensibly more important than the former. > > > > > > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > >> Thanks for raising this issue. Just a few points about the other side of >> the argument: >> >> It?s well known that things that are free are not valued (by humans) as >> much as those that cost something, even a small amount. For example, see >> Dan Ariely?s ?Predictably Irrational? which presents the results of several >> experiments that demonstrate this. As a relevant case: Free MOOCs tended to >> have lots of ?sign ups? but far fewer attendees, and even fewer completers. >> >> As such, if the goal is to have engaged attendees, trying to come closer >> to the experience of traditional conferences, it might make sense to charge >> something, even a small amount like $25, for at least some of the >> population. This population might be people who have lots of social capital >> already, and are generally busy, so they are more likely to blow off the >> conference if they paid nothing for signing up. Such people might be those >> that more junior attendees wish to meet. >> >> I note that engaged attendance was a goal when we had in-person >> conferences, so I don?t see why we?d want to drop it now. Indeed, if people >> don?t want to be engaged the videos will be available for free, afterward. >> >> Beyond the modest fees to run an online conference, which Talia mentions, >> conference registration payments serve other purposes. Any surplus goes to >> SIGPLAN, which turns around this surplus as good works, e.g., paying the >> open access fees for PACMPL, which ICFP benefits from. It also makes >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >> >> Corporate sponsors can indeed pay some costs, but they also have >> downsides. We are finding that many sponsors are not interested in >> necessarily giving that much, and some are starting to make demands on how >> the conference is run for their modest donation. This is a slippery slope >> that the SIGPLAN EC is trying to avoid. >> >> Given that PLDI was completely free and ICFP followed a progressive fee >> schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, >> to see how the registration fee affected attendance. >> >> Thanks, >> -Mike >> >> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer >> wrote: >> >>> [ The Types Forum, >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >>> >>> >>> >>> I don't know about PLDI, but there are some costs associated with online >>> >>> events. For example, automatic captioning software is still not very good >>> >>> (Google's always turns "proofs" into "fruits" for me). Live captioning is >>> >>> really expensive! But it's also hugely important for disability >>> >>> accessibility. >>> >>> >>> >>> For students, ICFP was essentially free. I do agree that in principle, >>> >>> online conferences should be free, and online components of hybrid >>> >>> conferences should be free or strongly discounted. In practice, though, I >>> >>> do think that will mean finding sponsors for hidden costs that really are >>> >>> necessary. >>> >>> >>> >>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >>> gabriel.scherer at gmail.com> >>> >>> wrote: >>> >>> >>> >>> > [ The Types Forum, >>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>> >>> > ] >>> >>> > >>> >>> > Dear types-list, >>> >>> > >>> >>> > Going on a tangent from Flavien's earlier post: I really think that >>> online >>> >>> > conferences should be free. >>> >>> > >>> >>> > Several conferences (PLDI for example) managed to run free-of-charge >>> since >>> >>> > the pandemic started, and they reported broader attendance and a strong >>> >>> > diversity of attendants, which sounds great. I don't think we can >>> achieve >>> >>> > this with for-pay online conferences. >>> >>> > >>> >>> > ICFP is coming up shortly with a $100 registration price tag, and I >>> did not >>> >>> > register. >>> >>> > >>> >>> > I'm aware that running a large virtual conference requires computing >>> >>> > resources that do have a cost. For PLDI for example, the report only >>> says >>> >>> > that the cost was covered by industrial sponsors. Are numbers publicly >>> >>> > available on the cost of running a virtual conference? Note that if we >>> >>> > managed to run a conference on free software, I'm sure that >>> institutions >>> >>> > and volunteers could be convinced to help hosting and monitoring the >>> >>> > conference services during the event. >>> >>> > >>> >>> From nltd at nguyentito.eu Sun Sep 13 20:12:52 2020 From: nltd at nguyentito.eu (=?UTF-8?Q?Nguy=E1=BB=85n_L=C3=AA_Th=C3=A0nh_D=C5=A9ng?=) Date: Mon, 14 Sep 2020 02:12:52 +0200 Subject: [TYPES] Pledge in favor of open access Message-ID: <1be3e746-f602-4d0b-aa06-72c24dbc1b92@www.fastmail.com> Dear Types list, Since there have been some mentions of publication practices in the recent discussions on conferences, I would like to advertize the following pledge that you can sign in favor of open access (which I don't think has appeared on this mailing-list yet): https://nofreeviewnoreview.org/ To summarize: "we will avoid serving as peer reviewers for venues that do not make publicly available the research that we review. Instead, we will give priority to open-access venues in how we allocate our reviewing time and organizational efforts." NB: The pledge doesn't say anything about declining to submit as an author to closed-access venues. This is a deliberate choice, discussed in the FAQ: https://nofreeviewnoreview.org/faq#authors https://nofreeviewnoreview.org/faq#decline_publish Of course, someone who signs this pledge is entirely free to also commit to only publish in OA journals/conferences if they think it lends more credibility to their stance (personally this is what I do). Best, L? Th?nh D?ng (Tito) Nguy?n From gabriel.scherer at gmail.com Sun Sep 27 12:23:26 2020 From: gabriel.scherer at gmail.com (Gabriel Scherer) Date: Sun, 27 Sep 2020 18:23:26 +0200 Subject: [TYPES] online conferences should be free (was: global debriefing over our virtual experience of conferences) In-Reply-To: References: Message-ID: Dear Jens Palsberg (and types-list), In my previous email, I asked for budget summaries for past online SIGPLAN conferences (PLDI and ICFP, for example; more data is always welcome). Could SIGPLAN do it? One inspiring example of financial transparency would be Andreas Zeller's budget summary for the ISSTA'16 conference (organized by SIGSOFT): https://andreas-zeller.info/2018/02/01/where-your-conference-fees-go-to.html On the SIGPLAN blog you mentioned costs and fees as the "elephant in the room" ( https://blog.sigplan.org/2020/09/15/virtual-conferences-and-sigplan/ ). We may have different perspectives looking at different parts of the elephant; but the "blind men and an elephant" parable relies on the blind men openly sharing their information. Best On Tue, Aug 25, 2020 at 10:24 PM Gabriel Scherer wrote: > On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: > >> Beyond the modest fees to run an online conference, which Talia mentions, >> conference registration payments serve other purposes. Any surplus goes to >> SIGPLAN, which turns around this surplus as good works, e.g., paying the >> open access fees for PACMPL, which ICFP benefits from. It also makes >> donations to CRA-W, OPLSS, etc. and provides scholarships for PLMW. >> > > On this point I have a simple request: could SIGPLAN release a budget > summary for ICFP and PLDI this year? To have an informed discussion on > online conference fees, we should know the amounts of the varied costs > (including the "good works"), how much money comes in from sponsoring, and > how much comes from conference fees. > > Since I raised the question of conference fees I have heard varied > explanations from various people, for example the idea that PLDI was > decided to have free registration before the actual conference-running > costs were known, and that the $100 fee for ICFP adjusts to cover direct > costs better. Your message rather suggests that the direct costs use only a > modest fraction of the $100 (or maybe they can be covered entirely by > sponsorship ?), but we cannot tell without any actual data. Others > suggested that making the conference free was maybe doable for flagship > conferences with an industrial presence such as ICFP or PLDI, but that this > expectation could be problematic for more theoretical conferences with less > sponsors. But then, FSCD and IJCAR reportedly ran fine without registration > costs. > > (I don't think this is an outlandish or surprising request, for example I > remember Adam Chlipala making exactly this request during a conference Town > Hall a few years ago.) > > On Tue, Aug 25, 2020 at 10:07 PM Gabriel Scherer < > gabriel.scherer at gmail.com> wrote: > >> Thanks Mike for bringing clear arguments in favor. >> >> There was an interesting discussion with Nicolai, Henning and Stefan on >> the argument of registration fees helping engagement. I have some short >> points to make on this topic: >> >> 1. This exact argument (price signals value) is given by some high-ranked >> universities in the English-speaking part of the world that charge tens of >> thousands of dollars of tuition fee and *could* relatively easily cover the >> corresponding costs on their endowment money. >> >> 2. Just like for tuition fees, the slope is slippery. Mike mentioned $25 >> (for non-students) to help engagement by making the conference feel >> valuable, but now ICFP cost $100; one could argue for POPL'21 registration >> fees of $1000 for non-students, to make it *very* valuable as a conference. >> (As long as the people making pricing decisions are getting reimbursed for >> their registration fees, I guess we could pull this off? >> >> However, I think that this discussion on engagement is somewhat of a >> distraction. Having conference where people participate actively is >> certainly a good thing, it is *nice*. But sharing our knowledge and results >> in the most open way possible is *a core tenet of our duty as researchers*. >> For me it is clear that we should find *other* ways to help engagement that >> do not raise the barrier to entry to our conference, because the later is >> sensibly more important than the former. >> >> >> >> >> >> On Sun, Aug 23, 2020 at 4:54 PM Michael Hicks wrote: >> >>> Thanks for raising this issue. Just a few points about the other side of >>> the argument: >>> >>> It?s well known that things that are free are not valued (by humans) as >>> much as those that cost something, even a small amount. For example, see >>> Dan Ariely?s ?Predictably Irrational? which presents the results of several >>> experiments that demonstrate this. As a relevant case: Free MOOCs tended to >>> have lots of ?sign ups? but far fewer attendees, and even fewer completers. >>> >>> As such, if the goal is to have engaged attendees, trying to come closer >>> to the experience of traditional conferences, it might make sense to charge >>> something, even a small amount like $25, for at least some of the >>> population. This population might be people who have lots of social capital >>> already, and are generally busy, so they are more likely to blow off the >>> conference if they paid nothing for signing up. Such people might be those >>> that more junior attendees wish to meet. >>> >>> I note that engaged attendance was a goal when we had in-person >>> conferences, so I don?t see why we?d want to drop it now. Indeed, if people >>> don?t want to be engaged the videos will be available for free, afterward. >>> >>> Beyond the modest fees to run an online conference, which Talia >>> mentions, conference registration payments serve other purposes. Any >>> surplus goes to SIGPLAN, which turns around this surplus as good works, >>> e.g., paying the open access fees for PACMPL, which ICFP benefits from. It >>> also makes donations to CRA-W, OPLSS, etc. and provides scholarships for >>> PLMW. >>> >>> Corporate sponsors can indeed pay some costs, but they also have >>> downsides. We are finding that many sponsors are not interested in >>> necessarily giving that much, and some are starting to make demands on how >>> the conference is run for their modest donation. This is a slippery slope >>> that the SIGPLAN EC is trying to avoid. >>> >>> Given that PLDI was completely free and ICFP followed a progressive fee >>> schedule, I?ll be curious to compare the ICFP outbrief with that of PLDI?s, >>> to see how the registration fee affected attendance. >>> >>> Thanks, >>> -Mike >>> >>> On Sun, Aug 23, 2020 at 10:25 AM Talia Ringer >>> wrote: >>> >>>> [ The Types Forum, >>>> http://lists.seas.upenn.edu/mailman/listinfo/types-list ] >>>> >>>> >>>> >>>> I don't know about PLDI, but there are some costs associated with online >>>> >>>> events. For example, automatic captioning software is still not very >>>> good >>>> >>>> (Google's always turns "proofs" into "fruits" for me). Live captioning >>>> is >>>> >>>> really expensive! But it's also hugely important for disability >>>> >>>> accessibility. >>>> >>>> >>>> >>>> For students, ICFP was essentially free. I do agree that in principle, >>>> >>>> online conferences should be free, and online components of hybrid >>>> >>>> conferences should be free or strongly discounted. In practice, though, >>>> I >>>> >>>> do think that will mean finding sponsors for hidden costs that really >>>> are >>>> >>>> necessary. >>>> >>>> >>>> >>>> On Sun, Aug 23, 2020 at 7:07 AM Gabriel Scherer < >>>> gabriel.scherer at gmail.com> >>>> >>>> wrote: >>>> >>>> >>>> >>>> > [ The Types Forum, >>>> http://lists.seas.upenn.edu/mailman/listinfo/types-list >>>> >>>> > ] >>>> >>>> > >>>> >>>> > Dear types-list, >>>> >>>> > >>>> >>>> > Going on a tangent from Flavien's earlier post: I really think that >>>> online >>>> >>>> > conferences should be free. >>>> >>>> > >>>> >>>> > Several conferences (PLDI for example) managed to run free-of-charge >>>> since >>>> >>>> > the pandemic started, and they reported broader attendance and a >>>> strong >>>> >>>> > diversity of attendants, which sounds great. I don't think we can >>>> achieve >>>> >>>> > this with for-pay online conferences. >>>> >>>> > >>>> >>>> > ICFP is coming up shortly with a $100 registration price tag, and I >>>> did not >>>> >>>> > register. >>>> >>>> > >>>> >>>> > I'm aware that running a large virtual conference requires computing >>>> >>>> > resources that do have a cost. For PLDI for example, the report only >>>> says >>>> >>>> > that the cost was covered by industrial sponsors. Are numbers publicly >>>> >>>> > available on the cost of running a virtual conference? Note that if we >>>> >>>> > managed to run a conference on free software, I'm sure that >>>> institutions >>>> >>>> > and volunteers could be convinced to help hosting and monitoring the >>>> >>>> > conference services during the event. >>>> >>>> > >>>> >>>> From a.w.laarman at liacs.leidenuniv.nl Thu Nov 5 11:41:43 2020 From: a.w.laarman at liacs.leidenuniv.nl (Laarman, A.W.) Date: Thu, 5 Nov 2020 16:41:43 +0000 Subject: [TYPES] Call for Papers - SPIN 2021 Message-ID: <5A07EDEA-95A7-4FC6-AD31-8B5EA4C9DAEC@liacs.leidenuniv.nl> ******************************************************************************* Call for Papers SPIN 2021 International Symposium on Model Checking of Software July 14-15, 2021 Aarhus, Denmark Conference website: https://conf.researchr.org/home/spin-2021 Submission link: https://easychair.org/conferences/?conf=spin20210 The 27th edition of the SPIN symposium aims at bringing together researchers and practitioners interested in automated tool-based techniques for the analysis of software as well as models of software, for the purpose of verification and validation. The symposium specifically focuses on concurrent software but does not exclude the analysis of sequential software. Submissions are solicited on theoretical results, novel algorithms (classical and quantum), tool development, including for modern hardware (parallel and distributed), and empirical evaluation. Topics of interest include, but are not limited to: * Formal verification techniques for automated analysis of software * Formal analysis for modeling languages, such as UML/state charts * Formal specification languages, temporal logic, design-by-contract * Model checking * Automated theorem proving, including SAT and SMT * Verifying compilers * Abstraction and symbolic execution techniques * Static analysis and abstract interpretation * Combination of verification techniques * Modular and compositional verification techniques * Verification of timed and probabilistic systems * Automated testing using advanced analysis techniques * Combination of static and dynamic analyses * Derivation of specifications, test cases, or other useful material via formal analysis * Case studies of interesting systems or with interesting results * Engineering and implementation of software verification and analysis tools * Benchmark and comparative studies for formal verification and analysis tools * Formal methods of education and training * Insightful surveys or historical accounts on topics of relevance to the symposium * Relevant tools and algorithms for modern hardware, e.g.: parallel, GPU, TPU, FPGA, cloud, and quantum Important Dates Submission: March 1, 2021 Notification: May 14, 2021 Camera-ready: June 1, 2021 Conference: July 14-15, 2021 Submission Guidelines The proceedings of SPIN 2020 will be published in Springer's Lecture Notes in Computer Science series. Submissions should adhere to the LNCS format: LNCS Information for Authors With the exception of survey and history papers, the papers should contain original work that has not been submitted or accepted for publication elsewhere. We are soliciting three categories of papers: * Full Research / Tool Papers describing fully developed work and complete results (16 pages - references are not included in this limit); * Short Papers presenting tools, technology, experiences with lessons learned, new ideas, work in progress with preliminary results, and novel contributions to formal methods (6 pages - references are not included in this limit). * Tool Demo Papers presenting the foundations, capabilities, application domains and relevant examples using the tools, with a clear description of what is expected to be shown in a live demonstration (4 pages to describe the tool foundations, features and use examples, plus an appendix explaining the content of the demo). Papers should be submitted via the EasyChair SPIN 2021 submission website: https://easychair.org/conferences/?conf=spin20210 All papers that conform to submission guidelines will be peer-reviewed by members of the program committee. Submissions will be evaluated on the basis of originality, the importance of contribution, soundness, evaluation, quality of presentation, and appropriate comparison to related work. At least one author of each accepted paper must attend the symposium and present the paper. STTT A selection of papers will be invited to a special issue of the International Journal on Software Tools for Technology Transfer (STTT). Program Committee Members Jiri Barnat, Masaryk University Maurice H. ter Beek, ISTI-CNR Tom van Dijk, University of Twente Vedran Dunjko, Leiden University Stefan Edelkamp, University of Koblenz Grigory Fedyukovich, Princeton University Henri Hansen, Tampere University of Technology Arnd Hartmanns, University of Twente Gerard Holzmann, Nimble research Antti Hyvarinen, Universia della Svizzera italiana Nils Jansen, Radboud University Peter Gjol Jensen, Aalborg Univesity Sung-Shik Jongmans, Open University, CWI Jeroen Keiren, Eindhoven University of Technology Igor Konnov, Informal Systems Inc Alberto Lluch Lafuente, Technical University of Denmark Kuldeep Meel, National University of Singapore Alica Miller, University of Glascow Sergio Mover, Ecole Polytechnique Rajagopal Nagarajan, The University of Warwick Doron Peled, Bar Ilan University Tatjana Petrov, University of Konstanz Jaco van de Pol, Aarhus University Stephen Siegel, University of Delaware Carsten Sinz, Karlsruhe Institute of Technology Jiri Srba, Aalborg University Michael Tautschnig, Amazon Web Services Yann Thierry-Mieg, Sorbonne University - LIP6 Yakir Vizel, Technion Georg Weissenbacher, Vienna University of Technology Anton Wijs, Eindhoven University of Technology Organizing committee Alfons Laarman, Leiden University Ana Sokolova, University of Salzburg Venue Established in 1928, Aarhus University has since developed into a major Danish university with a strong international reputation with approximately 38,000 students and 8,000 members of staff. Located in beautiful nature and wildlife areas, Aarhus is the second largest city of Denmark, with a rich trade and ancient Viking history. Contact All questions about submissions should be emailed to a.w.laarman at liacs.leidenuniv.nl. From rdbrown0au at gmail.com Mon Nov 9 06:42:56 2020 From: rdbrown0au at gmail.com (Rodney Brown) Date: Mon, 9 Nov 2020 22:42:56 +1100 Subject: [TYPES] Types for Units-of-Measure Message-ID: Reading ACM SigPlan Notices in the 80s, the papers on adding Units-of-Measure to programming languages for type-checking seemed sensible - especially after reading Comp.risks in Software Engineering Notes. More recently I read George Hart's 1995 book "Multidimensional Analysis: ..." which provided the abstract algebra behind my vague intuitive understanding of Dimensional correctness, sketched an implementation and described the suprising (to me) result that some problems in science and engineering need matrices that aren't simply dimensionally uniform. Using google scholar for literature searches on the subject, led to Kennedy's 1996 Thesis which describes adding Units of Measure to Standard ML. His later work adds this in F#. His thesis doesn't reference Hart though. Last year Griffioen's thesis extended Kennedy's Hindley?Milner type system to support Hart's matrices. I hope drawing attention to this may help bring Unit-of-Measure type checking to languages commonly used. en.wikipedia.org/wiki/User:RDBrown/Prog_Lang_Dimensions (has the links) ??? Kennedy, Andrew J. (April 1996). Programming languages and dimensions (Phd). 391. University of Cambridge. ISSN 1476-2986. UCAM-CL-TR-391. ??????? Kennedy, A. (2010). "Types for Units-of-Measure: Theory and Practice". In Horv?th, Z.; Plasmeijer, R.; Zs?k, V. (eds.). Central European Functional Programming School. CEFP 2009. Lecture Notes in Computer Science. 6299. Springer. doi:10.1007/978-3-642-17685-2_8. ISBN 978-3-642-17684-5. F# implementation usage tutorial ??? Hart, G. (1994). "The Theory of Dimensioned Matrices". In Lewis, John G. (ed.). Proceedings of Fifth SIAM Conference on Applied Linear Algebra, Snowbird, Utah, June 1994. Society for Industrial and Applied Mathematics. pp. 186?190. ISBN 9780898713367. ??? Hart, George W. "Multidimensional Analysis". ? links to the above paper and the book of the same name, showing examples of the multidimensional linear algebra theorems. ??????? Hart, George W. (1995), Multidimensional Analysis: Algebras and Systems for Science and Engineering, Springer-Verlag, ISBN 978-0-387-94417-3 ??????????? ISBN 9781461286974 reprint at Springer ??? Griffioen, P. R. (September 2015). "Type inference for array programming with dimensioned vector spaces". IFL '15: Proceedings of the 27th Symposium on the Implementation and Application of Functional Programming Languages. pp. 1?12. doi:10.1145/2897336.2897341. ISBN 978-1-4503-4273-5. "... We extend arrays with units of measurement, and Hindley-Milner typing with a matrix type based on the algebraic structure of units of measurement in matrices that allows type inference up to dimensioned vector spaces. The inference is sound and complete and gives the most general type of any linear algebra expression. Experiments show that the explicit support for linear algebra increases type safety, and that it leads to a more functional and index-free style of programming. It refines the types for linear algebra operators significantly, while the use of arrays as general containers has to be replaced by other data structures. As validation the matrix type system is implemented in the functional matrix language Pacioli." ??? Griffioen, P. (2019). A unit-aware matrix language and its application in control and auditing (PDF) (Thesis). University of Amsterdam. hdl:11245.1/fd7be191-700f-4468-a329-4c8ecd9007ba. ??????? github.com/pgriffel/pacioli From francois.pottier at inria.fr Mon Nov 9 07:56:11 2020 From: francois.pottier at inria.fr (=?UTF-8?Q?Fran=c3=a7ois_Pottier?=) Date: Mon, 9 Nov 2020 13:56:11 +0100 Subject: [TYPES] Types for Units-of-Measure In-Reply-To: References: Message-ID: <7c8b90c6-5eb6-6bb1-3788-85608f5190d3@inria.fr> Hello, See also Jacques Garrigue and Dara Ly's work (in French): https://www.math.nagoya-u.ac.jp/~garrigue/papers/ocamldim.pdf Back in the time, type inference with units of measure was investigated by Jean Goubault, concurrently with Kennedy, AFAIK. E.g., also in French: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.40.2594&rep=rep1&type=pdf -- Fran?ois Pottier francois.pottier at inria.fr http://cambium.inria.fr/~fpottier/ From reuben.rowe at rhul.ac.uk Mon Nov 9 09:11:34 2020 From: reuben.rowe at rhul.ac.uk (Reuben Rowe) Date: Mon, 9 Nov 2020 14:11:34 +0000 Subject: [TYPES] [EXT] Re: Types for Units-of-Measure In-Reply-To: <7c8b90c6-5eb6-6bb1-3788-85608f5190d3@inria.fr> References: <7c8b90c6-5eb6-6bb1-3788-85608f5190d3@inria.fr> Message-ID: <30a4acb8-40b0-b997-0644-57c493def4f3@rhul.ac.uk> Hi, I am also aware of the CamFort project, which applies this type of analysis to Fortran programs. https://camfort.github.io/ Reuben On 09/11/2020 12:56, Fran?ois Pottier wrote: > [ The Types Forum, > https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=VIHpUoSj5yhUtqcKK4tug7kGMy2eHuNsCX%2FoQrj241c%3D&reserved=0 > ] > > > Hello, > > See also Jacques Garrigue and Dara Ly's work (in French): > https://eur03.safelinks.protection.outlook.com/?url=https:%2F%2Fwww.math.nagoya-u.ac.jp%2F~garrigue%2Fpapers%2Focamldim.pdf&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=z9WKD6WiseuOUT%2B21LqaYhu0BSMyrbcw%2FmOtdXwn7Dk%3D&reserved=0 > > > Back in the time, type inference with units of measure was > investigated by Jean Goubault, concurrently with Kennedy, > AFAIK. > > E.g., also in French: > https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.40.2594%26rep%3Drep1%26type%3Dpdf&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=CckRSCgPaNxKG%2Fvou%2FfPLM8G4wrIOayVruZIm0qFcQU%3D&reserved=0 > > This email, its contents and any attachments are intended solely for the addressee and may contain confidential information. In certain circumstances, it may also be subject to legal privilege. Any unauthorised use, disclosure, or copying is not permitted. If you have received this email in error, please notify us and immediately and permanently delete it. Any views or opinions expressed in personal emails are solely those of the author and do not necessarily represent those of Royal Holloway, University of London. It is your responsibility to ensure that this email and any attachments are virus free. From c.grelck at uva.nl Mon Nov 16 15:08:20 2020 From: c.grelck at uva.nl (Clemens Grelck) Date: Mon, 16 Nov 2020 21:08:20 +0100 Subject: [TYPES] Types for Units-of-Measure In-Reply-To: <30a4acb8-40b0-b997-0644-57c493def4f3@rhul.ac.uk> References: <7c8b90c6-5eb6-6bb1-3788-85608f5190d3@inria.fr> <30a4acb8-40b0-b997-0644-57c493def4f3@rhul.ac.uk> Message-ID: <8b267cad-0b0e-c046-57fc-3564d7ccd1bf@gmail.com> Dear all, Paul Cockshott's Vector Pascal language might also be of interest: http://www.dcs.gla.ac.uk/~wpc/reports/compilers/compilerindex/x25.html (The website looks a bit deprecated though.) ? Clemens On 09.11.20 15:11, Reuben Rowe wrote: > [ The Types Forum, > http://lists.seas.upenn.edu/mailman/listinfo/types-list ] > > Hi, > > I am also aware of the CamFort project, which applies this type of > analysis to Fortran programs. > > ??? https://camfort.github.io/ > > Reuben > > On 09/11/2020 12:56, Fran?ois Pottier wrote: >> [ The Types Forum, >> https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Flists.seas.upenn.edu%2Fmailman%2Flistinfo%2Ftypes-list&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=VIHpUoSj5yhUtqcKK4tug7kGMy2eHuNsCX%2FoQrj241c%3D&reserved=0 >> >> ] >> >> >> Hello, >> >> See also Jacques Garrigue and Dara Ly's work (in French): >> https://eur03.safelinks.protection.outlook.com/?url=https:%2F%2Fwww.math.nagoya-u.ac.jp%2F~garrigue%2Fpapers%2Focamldim.pdf&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=z9WKD6WiseuOUT%2B21LqaYhu0BSMyrbcw%2FmOtdXwn7Dk%3D&reserved=0 >> >> >> >> Back in the time, type inference with units of measure was >> investigated by Jean Goubault, concurrently with Kennedy, >> AFAIK. >> >> E.g., also in French: >> https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.40.2594%26rep%3Drep1%26type%3Dpdf&data=04%7C01%7Creuben.rowe%40rhul.ac.uk%7C73be036f59174ab9eda508d884b437d9%7C2efd699a19224e69b601108008d28a2e%7C0%7C0%7C637405257139079382%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=CckRSCgPaNxKG%2Fvou%2FfPLM8G4wrIOayVruZIm0qFcQU%3D&reserved=0 >> >> >> > This email, its contents and any attachments are intended solely for > the addressee and may contain confidential information. In certain > circumstances, it may also be subject to legal privilege. Any > unauthorised use, disclosure, or copying is not permitted. If you have > received this email in error, please notify us and immediately and > permanently delete it. Any views or opinions expressed in personal > emails are solely those of the author and do not necessarily represent > those of Royal Holloway, University of London. It is your > responsibility to ensure that this email and any attachments are virus > free. -- ---------------------------------------------------------------------- Dr Clemens Grelck Science Park 904 Associate Professor 1098XH Amsterdam Programme Director MSc Software Engineering Netherlands University of Amsterdam Institute for Informatics T +31 (0) 20 525 8683 Systems and Networking Lab F +31 (0) 20 525 7490 Parallel Computing Systems Group Office C3.109 staff.fnwi.uva.nl/c.u.grelck ---------------------------------------------------------------------- From rae at richarde.dev Wed Dec 9 16:32:59 2020 From: rae at richarde.dev (Richard Eisenberg) Date: Wed, 9 Dec 2020 21:32:59 +0000 Subject: [TYPES] Optimizing data representation Message-ID: <010f0176496bf1e2-c63f2149-01ce-4b90-b6e1-c6acaa74727b-000000@us-east-2.amazonses.com> At last year's ICFP in Berlin, I was fortunate enough to head out for the evening with someone working on optimizing type representations. The work he described sounded amazing, but I have failed to recall any details helpful in finding the work to follow up on. The main idea is to analyze a type declaration to find more efficient ways of representation. I seem to think the analysis included both the type declaration and its use sites. For example, the tool could figure out that `data Nat = Zero | Succ Nat` should really be a number stored in binary, not a linked list with no information stored in the nodes. (I don't remember how it dealt with the unbounded nature of Nat vs the bounded nature of machine integers.) The analysis could also, if I recall, figure out that a list accessed by index is better represented as an array. Having performed this analysis, the tool could then transform the code to use the more efficient representation. Does anyone know of recent work in this direction? I seem to recall there was a publication in 2018 or 2019 about this all. It sounded very impressive, and I would like to know more. Thanks! Richard -=-=-=-=-=-=-=-=-=-=- Richard A. Eisenberg, PhD Principal Researcher at https://tweag.io https://richarde.dev/ From monnier at iro.umontreal.ca Thu Dec 10 13:49:23 2020 From: monnier at iro.umontreal.ca (Stefan Monnier) Date: Thu, 10 Dec 2020 13:49:23 -0500 Subject: [TYPES] Optimizing data representation In-Reply-To: <010f0176496bf1e2-c63f2149-01ce-4b90-b6e1-c6acaa74727b-000000@us-east-2.amazonses.com> (Richard Eisenberg's message of "Wed, 9 Dec 2020 21:32:59 +0000") References: <010f0176496bf1e2-c63f2149-01ce-4b90-b6e1-c6acaa74727b-000000@us-east-2.amazonses.com> Message-ID: > declaration and its use sites. For example, the tool could figure out that > `data Nat = Zero | Succ Nat` should really be a number stored in binary, not > a linked list with no information stored in the nodes. (I don't remember how > it dealt with the unbounded nature of Nat vs the bounded nature of machine > integers.) In most circumstances, the bounded nature of the memory means that the bounds on machine integers is no worse. > The analysis could also, if I recall, figure out that a list accessed > by index is better represented as an array. Having performed this > analysis, the tool could then transform the code to use the more > efficient representation. That sounds much more impressive than turning Nats into machine integers, indeed, I'm interested to hear what you find. Stefan From rae at richarde.dev Thu Dec 10 14:41:23 2020 From: rae at richarde.dev (Richard Eisenberg) Date: Thu, 10 Dec 2020 19:41:23 +0000 Subject: [TYPES] Optimizing data representation In-Reply-To: References: <010f0176496bf1e2-c63f2149-01ce-4b90-b6e1-c6acaa74727b-000000@us-east-2.amazonses.com> Message-ID: <010f01764e2c2117-dabb5eaa-9c06-4cf8-a46b-e61eebb70d9d-000000@us-east-2.amazonses.com> > On Dec 10, 2020, at 1:49 PM, Stefan Monnier wrote: > > That sounds much more impressive than turning Nats into machine > integers, indeed, I'm interested to hear what you find. > > I've been pointed to https://cozy.uwplse.org/, which I think is the project I was looking for. Thanks, TYPES! Richard -=-=-=-=-=-=-=-=-=-=- Richard A. Eisenberg, PhD Principal Researcher at https://tweag.io https://richarde.dev/ From jonathan.aldrich at cs.cmu.edu Mon Dec 21 22:43:01 2020 From: jonathan.aldrich at cs.cmu.edu (Jonathan Aldrich) Date: Mon, 21 Dec 2020 22:43:01 -0500 Subject: [TYPES] open letter: make POPL and future virtual conferences more inclusive Message-ID: Dear types members, If you've looked at the POPL 2021 schedule, you may have noticed that all POPL events are in a 5-hour time window that excludes anyone in East Asia, Australia, Oceania, etc. from synchronous participation. Notably, this is mostly people of color, and includes many areas where research in PL is nascent but growing, as well as some where PL research is strong. There's an open letter to POPL and SIGPLAN leadership asking that POPL 2021 and future virtual SIGPLAN conferences be inclusive of participants from around the world using time zone mirroring or other techniques: https://docs.google.com/document/d/16pfd5ljGu5urynHmYaW53DM2-rzryRJ0bisgc16IwKo PLDI surveyed attendees, and the survey results indicated that any future virtual PLDI conferences should "Commit to scheduling so that people from all timezones have approximately equal opportunity to engage with PLDI." See section 6.3 of https:/shttps:// arxiv.org/ftp/arxiv/papers/2007/2007.11686.pdf The SPLASH post-conference survey results also support this. So the community is fully supportive of geographic inclusion, but POPL organizers do not seem to be listening. I believe inclusion in all its forms is critical to the future of PL! If you agree, and think that POPL and future major virtual SIGPLAN conferences should therefore follow ICSE, ICFP, and SPLASH in using mirroring and other approaches to be inclusive of people around the world, I invite you to cosign the letter here: https://forms.gle/utgvTX2D82QKzWxm6 and encourage colleagues on social media and mailing lists to do likewise. Best, Jonathan Aldrich