Interesting. I'm on a project that switched from Mercurial to Git
recently. I don't have much of a sense of the relevant differences
but other people on the project seemed to prefer Git rather strongly.
Git does seem to faster, contrary to the measurements in that PEP.
From what I've seen by browsing the blogs of various large
communities, there always seems to be large numbers of Git advocates
(although "fanboys" might be a more accurate term in numerous cases)
ready to crowd out blog comments with single line insistences that
projects switch to Git. Indeed, I've even seen something resembling an
advocacy site for Git where there's a comparison to Mercurial and
Bazaar, although the benchmarks seem to be done with a certain amount
of honesty. Consequently, many projects seem to choose Git either on
the basis of what other people have supposedly done, or on the "hacker
cred" supposedly conferred through its association with Linus
Torvalds.
I've also heard various tales of Git usage where "strong antipathies"
would be a fairly accurate summary of user experiences, and this is
where such experiences can be contrasted with solid experiences with
other version control systems. And I've heard stories of "bait and
switch" with Git: "you can do XYZ with Git but not with ..." followed
by the discovery that you can't realistically do XYZ with Git, either.
Such advocacy makes me deeply cynical about Git.
To be fair, Git usability has apparently come some way since, say,
2006 when I first encountered Git and Mercurial, with the former being
recommended by the guys who develop itools (whose judgement I
generally trust) but with the caveat that one used the Cogito front-
end at that time, and the latter by the guys who develop MoinMoin
(whose judgement I also trust). The difference between the two DVCSs
as far as I have seen is that Mercurial has always been as usable as
it is today without additional tools. In addition, there seems to be
more honesty from the Mercurial community about what Mercurial can and
cannot support, as opposed to the disinformation cloud (or perhaps
uninformation cloud) that seems to surround Git.
As for the performance of these tools, you wouldn't think Mercurial
was written in Python (admittedly with some C) and Git was written in
C looking at various benchmarks. For what it's worth, this BDFL
decision is one I actually agree with.
Paul
