nr at eecs.harvard.edu
Fri Nov 23 19:45:49 EST 2007
> We've been using darcs for nearly 2 years now, I feel it's time for a
> retrospective and to think seriously about whether darcs is still the best
> option to continue with...
Your post about darcs was so thoughtful that I have almost nothing to add.
I did learn a bit about distributed revision control from an
interesting blog post and followups at http://tinyurl.com/2wktoq.
I would love to stay with darcs if it were faster and actually worked.
I don't see either of those things actually happening in the next five
years. I don't think there are enough good people working on the
underlying theory, the algorithms, or the code base. One (rather
snotty) way to state my opinion is this: I'll start to believe in
darcs's promises again once I see a POPL paper on the semantics.
git seems quite promising, primarily because it has the full power of
the linux kernel mafia behind it. It's fast, it has staying power,
and it will not be allowed to stay broken. I've read that the primary
disadvantage is that that it's very much optimized for a 'single
integrator' model, but for GHC that might be workable---I'm not sure I
know what the comment really means.
I'm no friend of darcs, but I would find it reasonable to continue to
limp along with darcs for another year or two in the hope that the
design space clarifies itself. Or maybe we could persuade Benjamin
Pierce that distributed version control is every bit as interesting as
file synchronization, and in a year or two we'd have a tool with a
real semantics :-)
More information about the Cvs-ghc