SRU policy for universe (was Re: Pending SRU Request; What to do?)
stefan.potyra at informatik.uni-erlangen.de
Tue Nov 20 12:38:58 GMT 2007
Am Dienstag, 20. November 2007 12:55 schrieb Martin Pitt:
> Stefan Potyra [2007-11-19 13:47 +0100]:
> > Well, the current SRU policy for universe was to have a very low entry
> > barrier for -proposed (aiming that archive admins should basically only
> > check the sanity of the version, due to a LP bug not being able to delete
> > packages from the -proposed queues)
> Side note: we can do that now.
> > and to have testing being performed as the main
> > reviewing measure. This means that no peer review for actual updates
> > would happen, unless archive admins would perform additional checks prior
> > to the migration from -proposed to -updates.
> > The change was somewhat drastic from very picky reviews by motu-sru to no
> > reviews for universe SRUs at all. The goal of this change was to
> > encourage more SRUs being done (as the strict policy we had before seemed
> > to have scared MOTUs away) and to improve the throughput of SRUs.
> There seem to be some fundamental philosophic differences here. You
> say that MOTUs want more SRUs, no entry barrier to -proposed, and
> -proposed being the play- and testing ground.
I don't think it is such a fundamental philosophic difference (apart
from -proposed being the play- and testing-ground), see my comments below.
Back then the number of SRUs for universe (compared to the # regressions) was
pretty limited, so that often enough backports were used as a measure to fix
regressions, which I guess shouldn't be the goal for a policy as well.
> My philosophy is that at least in main we imposed enough bureaucracy
> to *stop* people from doing non-essential SRUs, since:
> * We promise that it is a stable release and nothing changes unless
> we believe it is absolutely necessary *for existing installations*.
> This particularly means that we will not fix installation bugs,
> hardware support, etc, unless it is a major regression from a
> previous release.
That's what I believe we should be using SRUs for universe as well.
> * We want people to actually use -propose to give us feedback. If we
> put random untested stuff into it, then nobody will do so because
> they will risk breaking their machines every other day.
SRUs for universe (with the current policy) should never be untested atm,
however these are ...
> We don't let anything into proposed/main which has not been tested,
> peer reviewed, and acknowledged by ubuntu-sru.
... not peer reviewed.
> * Development efforts should go into the development release. There
> is no point in diverting a lot of effort into stables, since that
> effort will be missing in Hardy. This in turn will cause hardy
> universe to be worse than it could be and again bring up the need
> for hardy SRUs after release.
That makes complete sense to me. It's difficult to make a SRU policy for
universe that will aim at such a goal, as we've imho got a very large number
of bugs that would fall under the first point, but of course have an even
larger number of bugs to fix for the development release.
> As it stands, this philosophy will not change for main and restricted.
> It is of course possible to have your approach for
> universe/multiverse, but then we need to communicate this clearly and
> also change software-properties to provide separate checkboxes for
> enabling -proposed for main and universe (since we cannot recommend in
> good faith to turn on universe-proposed on production boxes).
Right, that's what I tried to write in the last mail already (but obviously
failed to express myself *g*).
As it stands, my opinion is to rather aim to make main and universe SRUs more
similar, at least for the semantics of -proposed. Unfortunately I didn't come
around to preparing some numbers about universe SRUs yet to look at if the
policy change did even have an impact on the number of actual universe SRUs.
I'll promise to do so either today or tomorrow and report back. Let's discuss
universe SRUs then further and try to come to a sane policy then, ok?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: not available
Url : https://lists.ubuntu.com/archives/ubuntu-devel/attachments/20071120/33305719/attachment.pgp
More information about the ubuntu-devel