[OLPC-devel] OLPC development project organization. Status calls? Other techniques?

Ivan Krstic krstic at fas.harvard.edu
Mon Jun 5 11:35:42 EDT 2006


David Woodhouse wrote:
> That'd be interesting -- I haven't really paid a huge amount of
> attention to Mercurial. I wouldn't be at all surprised if it's
> _slightly_ nicer than git in some esoteric way

The general thinking is that Mercurial is portable while git isn't, and
git's learning curve and default porcelain are sucky, while Mercurial is
very friendly.

OpenSolaris and, more surprisingly, Xen chose Mercurial over git. Ubuntu
and a host of other projects are running on top of the
Canonical-sponsored Bazaar-NG, and Mark Shuttleworth recently brought
out developers from both VCSes to do a combined sprint and see if they
can merge. It's extremely likely that whatever comes out of that
cooperation will soon become the dominant DVCS outside of the kernel
community.

As it stands, I doubt we'll have anyone developing OLPC applications on
Windows, so the portability is a non-issue. We then come down to
learning curve, overall simplicity, and possibly hackability. It'd be
helpful if someone could pitch in here, review the following two
overviews, and let us know if git seems as much more difficult to use as
it's usually portrayed. I'm very comfortable with both, so I can't judge
this fairly.

Tutorials/overviews:
 git -- http://wiki.sourcemage.org/Git_Guide
 hg  -- http://www.selenic.com/mercurial/wiki/index.cgi/QuickStart

> What we're looking for, however, is an _overriding_ reason why we need
> to double the number of version control systems we deal with; given that
> we have to use git for kernel stuff anyway.

I think I'm coming to agree. Particularly since there are some real
niceties that git offers us as far as branch/repo housekeeping goes, e.g.:
 http://www.kernel.org/git/?p=git/git.git;a=blob;f=Documentation/howto/update-hook-example.txt

Perhaps the best thing to do at this point is to go with git unless
someone screams about it being too complicated, and then possibly keep
our eyes open towards the bzr/hg combination in the future.

Thoughts?

-- 
Ivan Krstic <krstic at fas.harvard.edu> | GPG: 0x147C722D



More information about the Devel mailing list