How can we ensure Bazaar (bzr) remains active?
Stephen J. Turnbull
stephen at xemacs.org
Mon Sep 21 09:54:23 UTC 2015
Sebastien Alaiwan writes:
> I think the vast majority of people using git and "loving it"
> directly came from Subversion.
Perhaps. However, I know a lot of people (no claims about majority or
even large minority) who like git because git exposes the DAG to user
manipulation, git's "theory of operation" (not the UI!) fits in their
heads, and git is not opinionated about rebasing or mainline focus.
All this makes hacking on git and on git repositories relatively easy
for both git hackers and knowledgeable git users who have messed up
their repos in some way. Unfortunately, Bazaar hasn't had the same
attraction for the kind of users who like to sharpen their own tools.
In fact, it's frequently said by Bazaar advocates that they love
Bazaar because they don't *have* to think about it, implying that they
*don't* think about it. It's the VCS "for the rest of us", as the
Apple commercials used to say. That attitude has to change, at least
for a larger subset of Bazaar users than are currently hacking on
I think before you start aiming at World Domination, Bazaar needs to
get back on track for its current audience. Which it's losing AFAICS:
Emacs a year or so ago, and now GNU Mailman in the last few months.
Unfortunately, the reasons are rather different, and distressing to
those who believe that Bazaar is much better than git. In Emacs, a
majority of the developers who expressed opinions greatly favored git
as a VCS; and among those who opposed the change, those who opposed
change, not git, were prominent. When it became clear that ESR would
do all actual work, and no core Emacs resources would be expended
(except hot air, of which there is an abundance on emacs-devel), the
vocal Bazaar advocates shrank to a fairly tiny minority. *None* were
willing to contribute to improving Bazaar itself (beyond bug reports
and maybe doc improvements, at least to my recollection) -- despite
calls from Richard Stallman himself for volunteers to improve a
program that was both the official GNU VCS and that used by Emacs.
In the case of Mailman, it was the availability of an open source
GitHub clone (GitLab) that closed the deal. There were some
infelicities in the Launchpad workflow (one complaint that got a lot
of play, though I don't know how important it "really" is, was that
our GSoC students from India often needed HTTPS write access to the
public repos and Launchpad requires SSH for write access), and a lot
of developers just preferred git. Even those who prefer Bazaar had
trouble getting worked up because they already use git, as way too
many projects are adopting git now. So the costs of changing are now
the cost of a new clone, not self-reeducation.
Better support from public repos like Launchpad and Savannah really
does seem to be an issue, too. While Python uses hg, not bzr, there
is a whiff of change in the air there, too, again toward git. While
the git fanboys' claim that git is more powerful than Mercurial have
been effectively rebutted, I think, the claim that GitHub (and GitLab)
is more awesome than hg.python.org or BitBucket is accepted by several
of the most important developers. Python won't move fast on this; I'm
just providing (weak-ish) evidence that complementary services matter.
> > We cannot (and should not) allow such an awesome DVCS to fall
> > into a state of being unused, ignored, and eventually
> > abandoned. Rather, let's improve it, spread it, and use it!
> Amen brother.
Despite my personal preference, I agree with the sentiment. Just be
realistic about how you go about it.
I'm not sure what needs to be done. Python 3 support is obviously
important in the long run, and may be helpful in providing an image of
"keeping up with the times." I agree that integrating "foreign"
repository support would be helpful, but that may be hard to do in a
way that preserves Bazaar's "look and feel" (specifically in managing
a well-defined mainline, in the presence of git's colocated branches).
Finally, nested projects are useful.
For attracting git users, more transparent support for colocated
branches would be helpful.
Again, I don't have a sense of how to prioritize those to preserve and
deepen the pool of Bazaar fans, but they deserve consideration IMO.
More information about the bazaar