OpenOffice 3.1

Adam Williamson awilliam at redhat.com
Tue May 12 16:41:50 UTC 2009


On Tue, 2009-05-12 at 18:54 +0300, Gilboa Davara wrote:
> On Tue, 2009-05-12 at 17:23 +0200, Matej Cepl wrote:
> > On 2009-05-12, 10:09 GMT, Gilboa Davara wrote:
> > > and RHEL 5.3 would have been left with Firefox 1.5 instead of 
> > > Firefox 3.0.
> > 
> > Please, keep RHEL out of this discussioni ... there are another 
> > reasons (mainly customers breaking-in our doors with RFE) for 
> > doing things we do, which are not applicable to Fedora.

> It wasn't my intention to implicate RHEL in any way.
> My point being that a "stable release policy" != "stagnant release
> policy". If the maintainer feels that a new version fixes major
> deficiencies in the existing version (beyond a simple bug fix) he should
> be allowed to push a major upgrade (given sufficient testing) even if it
> risk breaking existing installations.

I don't really agree. There's wiggle room in *everything*, of course -
Mandriva updates Firefox by major version jumps, when the old major goes
out of maintenance, because on balance that's safer than trying to
backport security fixes to a version of Firefox that's unsupported
upstream - but at the level of setting overall policy, I wouldn't be in
favour of that.

But, to re-iterate, since it's easy to lose this point, I'm talking
about *the theoretical 'stable' update source we don't yet have*, ONLY
in the case where we decide we want one (i.e. we care about Aunt Flo).

I would never have fed KDE 4.0 to Aunt Flo in the first place, and no
distribution which cares about Aunt Flo did, so the discussion is a bit
moot there. Mandriva, for instance, went to KDE 4 by default only with
KDE 4.1, and Mandriva certainly wouldn't consider shipping KDE 4.2 as a
'stable' update for KDE 4.1. Neither would Ubuntu. You can *get* KDE 4.2
for both Mandriva and Ubuntu releases that shipped with KDE 4.1, but
it's not released as a stable update, nor - IMO - should it be, *in a
distribution which uses stable updates* (which, right now, Fedora
isn't).

> ... And if you don't trust your maintainers to make such a decision -
> why the hell have you given them the right to package the software to
> begin with?

I don't think that follows at all. Updates policies, above all else,
need to be consistent, and if you say "well, we'll try to be quite
stable and safe, but ultimately it's up to the discretion of
maintainers", you will never get consistency, because all maintainers
are different. What Maintainer A considers a 'safe' upgrade will be
different to what Maintainer B considers safe. Hence the user experience
will be inconsistent. Effectively, the update process will only be as
safe as the most happy-go-lucky maintainer's position - even though all
other maintainers are more conservative. And that's not efficient,
because you get all the drawbacks of instability from the maintainers
who push whatever they like, and all the drawbacks of staleness from the
maintainers who act very conservatively.

This is why I say the only two policies that can really work optimally
are "minimal necessary changes to fix strictly identified bugs and
security flaws" or "update whatever you like". Either is valid, but both
have distinct implications for the user experience, so we need to pick
based on what user experience we want, and message that consistently so
that users know what to expect.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net




More information about the fedora-devel-list mailing list