Re: Git mirrors

[Óscar Fuentes]

Juanma Barranquero <address@hidden> writes:

> On Fri, Oct 14, 2011 at 01:26, Óscar Fuentes <address@hidden> wrote:
>
>> Why is it necessary to proclaim an official GNU Distributed Version
>> Control System (DVCS)? GNU has the technical goal of creating a Free
>> Unix-like OS. For that you strategically depend on a kernel, a compiler,
>> a linker, a shell... but a DVCS? Is it necessary to proclaim the
>> official GNU Solitaire card game?
>
> Are you seriously comparing Solitaire with a VCS?

Your irony detector is badly broken, Juanma. Call 112 before it is too
late :-)

> Most software
> projects are mainained with the help of a VCS, and if chosing one as
> "official", a DVCS seems more sensible than a non-distributed one. If
> you can do a meaningful comparison, compare having an "official VCS"
> with having an "official building tool", which is surely not
> necessay... Oh, wait. There's GNU make.

The existence of GNU make (or GNU bzr) does not immediately imply that
other tools with similar or overlapping functionality can't be more
appropriate for certain applications. AFAIR GNU is not free of
duplication. Let's suppose that the mercurial guys submit their tool to
GNU. So now what? The "official GNU DVCS" may be not as convenient as
other for certain purposes. (BTW, is there a "official GNU text editor?"
What can be?) Fact is: git, mercurial and bzr are not the same, in the
same sense that two linkers are the same as long as they are drop-ins
for each other. Actually, a DVCS is more similar to a programmer's
editor than to a linker. People don't feel attached to a linker. We all
know what happens with text editors.

>> (by promoting software that *could* be inferior to other
>> options hence increasing the risk of bad image or rejection;
>
> All software could be inferior. Most are, at one point or other of
> their lifecycle. CVS was the most techically advanced VCS, once.

So what? Is it good then to promote a package that is being largely
ignored by the community as if the words "official GNU whatever" were
magical? The developers out there are choosing their tools based on
their own best judgment (let's hope it is a sensible one) but GNU just
advertises a package as if it were the best possible option just because
someone filled some papers. Silly.

>> by making
>> harder to access or contribute to Free projects hosted by GNU;
>
> I really dislike this argument, because it means that people who wants
> to contribute to some free software project will avoid to do so
> because it does not use their tool of choice.

C'mon, Juanma. We are on emacs-devel. When the bug reporting system was
discussed, people threatened with not using the system or stopping their
contribution to Emacs altogether in case the system lacked X or required
Y (an email interface or a webforms interface, to be precise.) The same
thing just happened with the git mirror. And I somehow understand
that. Would you keep contributing your free time to a project if they
required something that makes you feel uncomfortable?

> But, by the same token,
> all projects should be written only in the most popular programming
> languages (C and Java, likely), and many do, but there are others.
> Let's face it, there are about five or so VCS with are mostly relevant
> to free software projects (CVS, Subversion, git, Bazaar and
> Mercurial), and to contribute you just have to learn a few commands.

And install the tool, and re-learn the commands every now and then
because you don't hack on Emacs every day, and cope with glitches
(apparent or real) on a unfamiliar tool, and use interfaces that makes
you cringe, specially when you remember how good are the ones available
next door...

[snip]

>> by
>> sending the message to other creators of Free Software that GNU is out
>> there to aggressively compete with them regardless of merit.)
>
> GNU software is out there to aggressively compete, and win the users,
> yes (at least the ones that care about freedom too). "Regardless of
> merit" is meaningless, because the users will chose the one which best
> fits their needs.

No if the user trusts GNU, and no if GNU promotes only the subset of
Free Software that was incorporated into GNU.

>> Have you read Karl Fogel's post on this same thread about how choosing
>> bzr for Emacs actually *damaged* bzr?
>
> It's funny that you use Karl's post to support your view, because he says:
>
>> It is *because* I support free software that I wish Emacs would switch
>> to Git on Savannah.  Doing so would be better for the cause.
>>
>> That's no slur on Bazaar.  It's just that Savannah clearly does not have
>> have the resources to support many different version control systems
>> well -- and as a result, we're not really helping Bazaar anyway.
>
> which is to say that Savannah is doing a better job of supporting git
> than Bazaar... That does not seem entirely compatible with the view
> that GNU is rejecting git or favoring Bazaar.

Now, we should ask the Savannah admins why bzr has problems while git
works rock-solid.

> As for Karl's comment, I think that the switch to Bazaar hurt that
> project PR-wise, yes, and the state of bzr support on Savannah had a
> part on it. But another, huge part, was the fact that bazaar was not,
> at the time, ready for a big project with a long history, like Emacs.

Precisely, Emacs did not a favor to bzr choosing it when it was not up
to the task. That's the typical consequence of short-sighted politics.

> But that was then. Currently, wanting to hack Emacs and not wanting to
> use Bazaar seems a bit childish IMO. Yes, you'd prefer to use git. I
> would prefer for Emacs to be written in Ada. I'll have to adapt. Can
> you?

I don't have write access nor have any planned contribution that
requires it, so I'm not the most relevant person to ask that
question. The git mirror serves me perfectly. And when I say "perfectly"
I mean "much better than bzr". If GNU decided to pull the plug on the
git mirror I'll be inconvenienced. Just that. Is it necessary?