Re: Switching to Git?

[Ludovic Courtès]

Hi,

Greg Troxel <address@hidden> writes:

> The more serious problem with guile is a lack of focus on timely, stable
> releases usable by people who want to integrate it.  Plus backwards
> compatibility so that people that do integrated it have very little
> grief, and the current slib mess.

(This is off-topic.)  I, for one, would like to help reach that goal,
but this is actually quite demanding.  For instance, fixing tiny build
problems on, say, Solaris does take time, unless the fellow reporter
provides a trustworthy patch (which rarely happens).  Fixing threading
issues is even more time-consuming, as you might guess.  I'm definitely
inclined towards producing more frequent stable releases, but you can't
expect timeliness from the bunch of volunteers that we are, sorry.

As for backwards compatibility, I don't think Guile has been doing such
a bad job.  The "SLIB mess" is actually no show-stopper for people that
use SLIB with Guile, AFAICS.

> I wonder if the 'decentralized development' notion is really consistent
> with the "papers, please" demand of FSF.  Have "assigned" projects used
> distributed CM systems?  How has that been handled?

Of course.  As I mentioned in my other mail, several GNU projects
switched to Git specifically (GnuTLS, libIDN, Coreutils, Gnulib,
Lilypond, etc.).

They handle copyright assignment the same way as you do when using CVS:
don't integrate other people's changes into the main line unless they
have done the paperwork.  Copyright assignment is an orthogonal issue.

Thanks,
Ludovic.