Re: Git mirrors

[Stephen J. Turnbull]

Juanma Barranquero writes:
 > On Fri, Oct 14, 2011 at 06:50, Stephen J. Turnbull <address@hidden> wrote:

 > > For sure.  And GNU now has two.  GNU Arch (since 2003), a definitely
 > > freedom-loving project.  And GNU Bazaar (just in time to be adopted by
 > > Emacs; coincidence?)  For heaven's sake, even the name "Bazaar" evokes
 > > open source ideals!
 > 
 > And?

The reason for cutting off discussion of what is good for Emacs (and
more generally, for GNU) was that Bazaar was a GNU project which
wasn't totally useless as a VCS in the context of Emacs traditional
workflow, although at the time performance was known to suck, which
took years to get fixed (mostly not not Bazaar's fault).  But Arch was
a tried and tested alternative, with a much longer history, a history
of actively supporting GNU goals, and an existing and well-tested
repository.

 > > I don't understand what you're trying to say.  Óscar is precisely
 > > arguing that there should be *no* "official" GNU VCS, because there
 > > are too many good ones out there.
 > 
 > I'm trying to say the same thing that Jambunathan K just said: that
 > the project's choice of a DVCS over another will only stop from
 > participating to those who weren't really inclined to do so in the
 > first place.

"Stop participating" is a strawman.  It's an exaggeration by folks who
claim it will happen, and now you turn around and again take it
absolutely and point out that it's greatly exaggerated, then
exaggerating that fact yourselves, into a claim of falsehood.  As
usual, the truth is somewhere in the middle, and IMO that truth
matters in practice.

The thing is, every minute spent on working around an inefficient
workflow is a minute that is not spent developing free software.  It
also changes the "price" of working on a particular project; if one
can use their VCS of choice in project A, and one they dislike
intensely in project B, project A will get more time than it otherwise
would, and project B less.  People will waste more time recovering
from mistakes in a VCS they don't like, etc.

Of course as you pointed out to Bastien, this door swings both ways.
So it's not an argument for switching to git, rather the reverse, as
many Emacs developers have come to like bzr AFAICT.

On the other hand, protecting bzr by discouraging use of git mirrors,
and git as the primary VCS by projects whose developers generally
prefer it, is costly to the GNU Project as a whole, by discouraging
(though not preventing) git fans from working on projects that conform
to the "use Bazaar" policy in spite of their own wishes.

All clear now?

 > > If users are choosing something other than GNU, and it's clear
 > > that GNU makes choices based on favoritism toward GNU-labeled
 > > projects, that makes the GNU recommendation meaningless as a
 > > signal of quality.
 > 
 > *Technical* quality, perhaps. But the recommendations are not just
 > technical,

Of course they're not, and in fact they are *primarily* nontechnical.
I'm simply saying that being a signal of technical quality is
something GNU *should* aim at, even though it *must* be a secondary
goal for GNU.

 > > It's already meaningless as a signal of the freedom of the
 > > software, since that is determined quite precisely by the
 > > license; no need for a GNU label.
 > 
 > That's not an argument against having a GNU DVCS,

Nobody claimed it is, just that there's no "pro" argument here.

 > it is an argument against having GNU in the first place.

Of course it's not!  While determining whether any one program is free
is relatively simple, it is a *huge* service to have a full system
delivered, and have confidence that everything in that system is free
software.

 > > It would be (economically) better if GNU developers making (currently)
 > > inferior software were encouraged to abandon their effort, and devote
 > > some of that time to improving the free rival(s)
 > 
 > Isn't that a recipe for monocultures?

That depends on the destination project(s).  In the case of OS kernels
and VCSes, I don't think the world at large would notice, or lose any
biodiversity if the development organizations of the HURD and Bazaar
simply disbanded but left their repos and tarballs archives in place.
Canonical customers would probably be disgusted, though.

 > Or are you suggesting that all XEmacs developers should abandon it,
 > sign papers and start hacking Emacs?

Hello, Eli!  We are now in the near vicinity of "ad hominem" (though
this *still* isn't an ad hominem).

No, I don't suggest that.  Abandon XEmacs, maybe, sign papers, maybe,
but I doubt that switching to Emacs would do anybody much good.

 > > But your analogy fails, because the problem here is not whether
 > > Óscar can *adapt* to Emacs' use of bzr.  He can, and he can use git
 > > (for developing Emacs) at the same time as bzr (for pushing his
 > > contributions) if he wants to.
 > 
 > Apparently, for Óscar is a problem.

Why do you keep ignoring what he writes?  Yes, he *wishes* Emacs used
git, but in this thread, he wants to know why GNU uses a policy that
appears to him to be counterproductive in a number of ways.

 > And, BTW, "failing to *conform*" is quite loaded, don't you think?

Sure.  It's accurate, though.  Richard hasn't said that using git is
Evil, nor that GNU projects *must* use bzr.  He'd just like to
"discourage" git use and "encourage" bzr use, whether or not the
individual projects and developers think that is good for them.  Doing
what the collective does at the expense of one's own interests is
accurately called "conforming".

 > > Richard's reluctance to express approval of this idea strikes me as
 > > going beyond *promoting* GNU Bazaar to *protecting* it.
 > 
 > And you're surprised that Richard is protective of GNU because...?

Of course Richard protects *GNU*, and nothing I have written should
give you the impression that I think that is anything but an
unqualified good.  Nor am I *surprised* that he protects individual
GNU projects; he has never been a fan of unbridled competition, and he
clearly favors somewhat tighter bridles than I do. :-)

I do think it's unfortunate that he makes a point of protecting GNU
Bazaar because (a) as a general principle I believe it harms GNU to
protect individual GNU projects from outside competition on merit,
just as any society with too many protected segments becomes weaker as
a whole, and (b) because Bazaar in particular is not an essential part
of the GNU system given the availability of several excellent free
alternatives, while the Bazaar project is hardly an enthusiastic
promoter of the GNU Project or its goals.  As far as their marketing
goes they emphasize cross-platform support (especially Windows
support), Launchpad, and Ubuntu, not GNU, and they almost never
mention software freedom (although of course they make a big deal that
their software is free software).