[Unison-hackers] On making times=true the default

[Greg Troxel]

Benjamin Pierce <bcpierce at cis.upenn.edu> writes:

> FWIW, I'm a little nervous about changing defaults on something so
> fundamental.
>
> E.g., we can be careful not to do it when using old archives where times
> have not been synced, but what if those archives get blown away and the
> user expects to be able to just run unison again to recreate them from
> scratch?  This is something some people might be quite surprised by.

Assuming we can implement automatic merge to older mtime, what are you
concerned about in terms of semantics and user expectations?

Do you think someone is relying on the mtimes of a replica where unison
rwote the file (as opposed to the user) being the time that it was
synced, instead of the most recent non-sync change?

If you do believe that people that want this exist, do you think they
are more than a tiny minority?   Would they be adequately served by a
caution in NEWS to set --times=false, to get this old semantics?

(I view lack of syncing mtimes as a bug, and I am guessing it was
inspired by Windows trouble.  But a historical perspective might be
helpful.)