[TYPES/announce] Two phase reviewing for POPL; a response

[Derek Dreyer]

> Precisely: the key word is IDEAS, not papers.
>
> The hiring and tenure committees should look for, and perhaps count,
> IDEAS, not papers, especially if at conferences. And the conferences
> should serve the purpose of disseminating new ideas, not of
> distributing medals.

Given that there have been several posts with something approaching
this sentiment, I would like to respectfully, and strongly, disagree.
POPL and other top conferences have been a place where many great
*ideas* in programming languages have been first set forth.  Yes,
there are definitely ways it can be improved---we can accept a larger
set of good papers, and we can ensure all papers are reviewed by
experts---but *fundamentally* the current system seems to me to work
amazingly well.

Conference papers are one of *the most effective* ways of transmitting
ideas that I know.  Their limited size forces the author(s) to be
concise, but at least in the case of POPL the size is large enough
that one has space to convey a significant contribution, with only
some of the technical details left to a tech report.  And the serious,
*selective* peer review forces authors to communicate their ideas
clearly.  Furthermore, conferences' regular frequency and deadlines
ensure that the ideas presented in conferences are often *not* in
their final stages of fruition.  They are not the final "polished"
word on subject X, but a well-explained intermediate result, which may
prove to be an important stepping stone to future results.  That is
why a good conference paper will clearly explain its limitations and
suggest ways of improving the results in future work.

So this claim that conference emphasize boring, polished papers over
good, novel ideas is rather mind-boggling to me.  It is true that
conferences do favor good ideas that are well-explained and can be
justified in the short term over good ideas that are somewhat sketchy
or can only be justified in the long term.  But this is a systemic
problem with research in general---i.e. nearly everything about the
"academic-industrial complex" favors concrete, short-term research
over fuzzy, long-term research---and turning conferences into
free-for-alls will not solve it.

Of course, journals provide an important, complementary, archival
role, but journal articles are not a good way of disseminating ideas
quickly.  They cannot scale to handle the number of submissions that
conferences handle, and journal articles give more information about a
piece of work than most people need to know in order to understand the
main ideas.  Journals do not replace conferences or vice versa.

So I agree with Harry Mairson that, while change may be in order,
what's needed is moderation.

> There are so many `papers' produced, that, right now, in computer
> `science', not even the authors read their own work. Thanks to cut &
> paste we have now a ratio of reads/writes < 1.

I don't know what you're talking about: I read way more papers than I
write, and I read way more conference papers than journal articles in
a given year, often because I am asked to review them!  Most of what I
learn about new ideas in PL is from reading conference papers.

Best,
Derek