LTS and release methodology

[Matt Zimmerman]

On Tue, Jul 08, 2008 at 04:13:23PM +0200, Pär Lidén wrote:
> 2008/7/8 Matt Zimmerman <mdz at ubuntu.com>:
> 
> > On Mon, Jul 07, 2008 at 01:00:00PM -0500, Luke L wrote:
> > > Ceteris paribus, regressions should have a higher priority than normal
> > > bugs. I totally agree.
> >
> > It's hard to argue with that, but again, I have to look at this
> > pragmatically.  It is very rarely possible to tell just by looking at a bug
> > whether it is a regression or not, and collecting this information can be
> > very time-consuming ("please boot from an older live CD and try it, then we
> > can decide on the importance of your bug").
> 
> 
> Yes, it might be tricky in some cases to figure this out, but quite often
> people say in the LP bugs "This worked in Feisty/Gutsy/Dapper". Maybe there
> could be some flag in LP to mark a bug as a regression? And those bugs would
> be required to get much more attention. It doesn't have to be more
> complicated than so.

There is a 'regression' tag, and we do try to prioritize these on an ad-hoc
basis, but understand that with such incomplete information, it's difficult
at best.

> Perhaps also an LTS release would be delayed if it has
> too many regressions? (I suppose a zero-tolerance to regressions would be
> too hard, and delaying the release for too long, but the kernel won't
> release a new version until all important regressions are fixed. IMHO,
> something similar should be the case for the whole Ubuntu releases also.)

This is already our policy; in fact, we delayed the first LTS release for
that reason.  8.04 released with some regressions, but these were considered
either a) not severe enough to warrant delaying the entire release, or b)
planned to be fixed with 8.04.1.

-- 
 - mdz