[darcs-devel] Re: using git directory cache code in darcs?

[Mark Stosberg]

On 2005-04-16, Junio C Hamano <junkio at cox.net> wrote:
>
>DR> 4) Would there be interest in creating a libgit?
>
> These are related.  I have seen some people interested in
> libifying it, and encapsulating those globals would naturally
> fall out of it.  My impression from the list however is that a
> lot more people are interested in the upper SCM layer than the
> git layer right now.  And git layer, although solid enough to
> host itself, is still slushy.  A couple of days ago dircache
> format was changed from host to network endian.  Last night
> Linus made another change to dircache format, which fortunately
> is upward compatible if you stay within pathnames shorter than
> 2^12 bytes ;-).  Another problem I see for somebody to pick up
> and start libifying things right now is that, although there is
> one central person on the SCM side (Petr Baudis), git layer is
> still fractured between Linus and Petr.
>
> Petr syncs with Linus often and he seems to be doing a good job
> at keeping track of public patches, but the git layer Linus
> works on does not have some patches Petr collected or wrote
> himself.  In time, a better coordination would emerge, of
> course, but the project is still young at this moment.  Stay
> tuned ;-).

On the darcs side, we need to clarify our priorities with respect to
get. Two potential directions include:

  1. Focus on getting 'git' technology into darcs.

  2. Focus on long term compatibility with 'libgit'.

I think "1" is more important, because it seems the end goal is an SCM
that can handle the Linux kernel, not a great library. However, this
focus could mean forking git, and losing compatibility. On the flip
side, the result may be darcs does in fact work well with large trees,
and kernel hackers like it. In that case, the 'git' code in darcs may
become more interesting to work on, and the primary one that people care
out. 

My take is that if it seems "good enough" right now, that's sufficient
reason to start trying to integrate it into darcs, even if it means
forking the code. 

(with the disclaimer that I'm not a C hacker!)

    Mark

-- 
http://mark.stosberg.com/