[darcs-devel] Darcs and git: plan of action

Juliusz Chroboczek Juliusz.Chroboczek at pps.jussieu.fr
Mon Apr 25 03:56:16 PDT 2005


> And now I am lost what the purpose of interfacing darcs and git should 
> be.  Is it using git within darcs?  Or importing from git into darcs?  
> Or exporting from darcs to git so that a re-import does not lose 
> information?  All of the above?  Or just some brainstorming?

We (the loose-knit group of people hacking at Darcs) are not necessari-
ly a homogeneous group.  David's motivations are likely to be very
different from mine.  Hopefully, we'll end up making a better Darcs
all around.

As I understand it, David's initial motivation was to use a git cache
as a pristine cache on steroids.  I happen not to share this motivation,
probably because I don't see how to make it sufficiently fast to be
worth the trouble; however, I'll be glad to collaborate on that when
there is a rough consensus based on working code.

My motivation is different.  Darcs has the best underlying model since
sliced bread, and the parts of the Darcs code that David is interested
in are pretty nice.  On the other hand, the parts that David, Ian and
a few prominent others are not interested in -- notably repository
access -- are in a pityful state.

The repository access code needs a complete restructuring, going
through abstract interfaces.  The trouble is -- it is difficult to
remodularise code ``in the abstract'', you have to abstract your code
for a good goal.

So to summarise, writing a second backend for Darcs is what we need in
order to remodularise the repository access code, which will make it
easier to maintain and easier to profile and optimise.  My personal
choice would have been to write an Arch backend (and I've been
thinking about doing that for quite some time), but since David seems
motivated with the idea of integrating Git into Darcs, why not use
that as an excuse.

                                        Juliusz





More information about the darcs-devel mailing list