reviving Fedora Legacy

[Patrice Dumas]

On Mon, Oct 13, 2008 at 09:23:42AM -0700, Jesse Keating wrote:
> On Mon, 2008-10-13 at 16:23 +0200, Ralf Corsepius wrote:
> > 
> > In a nutshell:
> > * Lift acls from all discontinued Fedoras
> > * Keep the buildinfrastructure alive.
> > * Allow fedora maintainers to fix bugs in discontinued Fedoras.
> 
> I think the only way I'd vote for this to happen is if it was setup in
> thusly:
> 
> At the point of official EOL (IE the 13~ month mark) PackageKit is
> notified in such a way so that in it's attempts to fetch new updates
> from existing repos it warns users about the EOL and explains to them
> the situation.  It can offer a few different options:
> 
> A) Upgrade (via pre-upgrade) to a newer version of Fedora.
> 
> B) Configure itself for a new set of repos that are the product of
> post-eol updates by maintainers, giving clear information to the user
> that there is no set lifespan, no set package list, no set expectation,
> no nothing.
> 
> C) Do nothing.
> 
> Users of EOL releases would have to make a conscious choice to jump on
> the post-eol update world, so that they are given a chance to understand
> what it is they would be consuming.  I would not want people to wind up
> there without any prior knowledge and properly set expectation of what
> they are getting, and I would want them to be urged to upgrade if at all
> possible.

Agreed. In fact I think that we could start something even without B)
available in packagekit, but only with a manual change in the yum config
files. But packagekit pointing to the post-eol repo would be even
better.

> Furthermore I would want to prevent maintainers who do not wish to
> participate in the post-EOL update world from getting bugzilla reports
> by people using their software in EOL releases.

My proposal is to do a mass orphaning of EOLed branches, such that a
maintainer has to actively step in. But I can't see a way to avoid bugs
labelled against a non EOLed branch to reach the current maintainer. If
a packager maintains a EOLed branch I also thing that it should be
mandatory (and also logical!) to be in watchbugzilla (if the maintainer 
agrees, of course) to do bug triaging, but for those that are orphaned
in the EOLed branch I can't see waht to do for wrongly labelled bugs.

> Finally, I would prefer if we had our automated signing system in place
> prior to this, but I'm not going to make this a blocker.
> 
> Maintainer teams of various systems such as Bodhi and pkgdb and koji and
> mirrormanager et al would also need to give their explicit approval and
> add any further adjustments to any such plan.

Of course. At least for me, I think that this idea is not worth pursuing
if the cost in term of infras manpower or even hardware and the like is
not almost nill (except for volunteers really interested in that).

--
Pat