Re: On the subject of Git, Bazaar, and the future of Emacs development

[Stephen J. Turnbull]

Giorgos Keramidas writes:

 > Of course.  I do not presume to say that Mercurial is going to be around
 > for more than $FOO,

As a program, CVS is still "around" and someone as authoritative (in
Emacs) as Stefan has had good words to say for it. :-)  Heck, I still
use RCS in some contexts.

The right question is "how much longer will CVS *the project* be
around?"  The answer to that is that that project died years ago.  git
seems to be moving much faster than Mercurial now, although Mercurial
development is far from "stopped".

Regarding git vs. hg from the viewpoint of Emacs:

The senior Python developers are almost all of the school that VC is a
necessary nuisance, so they don't display much interest in the VC as
long as their workflow continues smoothly.  For various reasons, their
workflow is much heavier weight than any VC is, so VC features are a
small consideration to them.  They probably won't change again until
that future generation of VCSes has as great advantage over Mercurial
as Mercurial (and other DVCSes) had over CVS 5 years ago.  (Really
good subtree/submodule support, or really good bidirectional merge
support would probably do it.)  Nor is there a strong grass-roots
movement to replace Mercurial, or voices demanding more features in
the VCS.

I believe that Emacs is different.  Although some senior developers
(rms, eliz, and Handa-san prominent among them) argued at the time for
a very conservative approach, as much like CVS as possible, others
have been active in DVCS for a long time (eg, Stefan and Miles have
been fiddling with other VCSes since Tom Lord's Arch was a collection
of bash scripts), and Eli himself has gone well beyond "minimal" use
of bzr.  There may be as many people using git to manage their Emacs
branches as there are bzr users, and everybody is agreed that the
current state of Bazaar is unsatisfactory.  To summarize, Emacs
developers as a group are pretty sensitive to improvements in the VCS,
and therefore it would be "nice" if they could have the leading VCS
most of the time.

It is my opinion that the architecture of git (including the plethora
of plumbing commands that people seem to love to hate) makes it the
odds-on favorite for the role of "leading VCS", more than Mercurial.
The rapid development of "cloud" implementations of git like GitHub
may be a hindrance from Emacs' point of view, though, because they
clearly decrease the pressure for improvements in git's CLI.

Caveat: rms doesn't consider any of that relevant at this point.