[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Moving away from (unsupportable) FusionForge on Alioth?



On Sun, 2017-05-14 at 10:49 -0700, Sean Whitton wrote:
> Hello,
> 
> On Sun, May 14, 2017 at 04:53:18PM +0800, Boyuan Yang wrote:
> > As a result, I'm writing to suggest we find an answer to such a problem soon.
> > Migration to Jessie or Stretch with new FusionForge version might be possible. 
> > Or we should just drop outdated FusionForge and move to some modern platforms 
> > like GitLab (with an alternated workflow possibly).
> > 
> > There are much room for discussion but we should start evaluation
> > without 
> > delay, since migration would take much time and the time left is
> > pretty 
> > limited.
> 
> One possibility that I don't believe has yet been raised is a minimal
> git hosting tool, like gitolite.  This is stable, supported software,
> and there is a good Debian package in the archive.
> 
> Alitoh is 90% simple git hosting, 5% managing push access to git repos,
> and 5% mailing lists.

Let's say VCS hosting, not git hosting.

> Alioth lists could be moved to lists.d.o.  As for
> access control, the current situation with -guest accounts must be
> rewritten anyway, because it is tied to FusionForge.  We could extend
> sso.d.o to handle registering -guest accounts, and then use some scripts
> to make all sso.d.o users visible to gitolite.

Would gitolite be able to support the most popular types of git hooks
(e.g. mail and IRC notifications for pushes)?

> Perhaps it is simply naïve to think that a piece of software as simple
> as gitolite could serve our needs.  However, one of the main blockers
> that keeps coming up in these threads is that many of us are very uneasy
> about monolithic web services like gitlab and pagure.  I suspect that a
> big part of the worry is that we'll be in a similar situation to where
> we find ourselves with FusionForge, a few years down the line.  So it is
> worth discussing alternatives to another monolithic web service.

There is no free lunch.  The more we want to be able to control our
code hosting, the more time Debian people will have to spend on it over
the long term.  I don't think it makes much difference whether the
implementation is monolithic or put together from more independent
pieces.

> (None of this helps with non-git repos on alioth, but neither do gitlab
> or pagure.)

Here's a tally of live packaging repositories hosted on Alioth, based
on Vcs fields for sources in unstable:

$ for type in arch bzr cvs darcs git hg mtn svn; do
>     printf '%s: ' $type
>     grep-dctrl -FVcs-$type -sPackage 'debian.org' /var/lib/apt/lists/httpredir.debian.org_debian_dists_unstable_*_Sources \
>     | wc -l
> done \
> | sort -k2 -nr
git: 18907
svn: 2377
bzr: 71
hg: 27
darcs: 22
arch: 7
cvs: 2
mtn: 0

It looks like git hosting would cover ~90% and git+svn would cover
>99%.

Ben.

-- 
Ben Hutchings
Life is what happens to you while you're busy making other plans.
                                                               - John
Lennon

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: