Deep Freeze coming for Fedora 7 (and cvs branching coming too)

[Ralf Corsepius]

On Tue, 2007-05-15 at 13:22 -0400, Jesse Keating wrote:
> On Tuesday 15 May 2007 13:15:07 Michael Schwendt wrote:
> > Eh? What has this comment to do here? Or have you realised that too many
> > users are burnt by the frequent updates and upgrades that are pushed out
> > to Fedora Core 5 and Fedora Core 6?
AFAICT, most of them wandered off because of updates misbehaving, due to
poor packaging, broken mirrors and yum suckage, not because of packages
misbehaving at run-time.

>  That is the #1 complaint by users who
> > have wandered off to Ubuntu, for example. Released Updates which break
> > something badly. Wrapped into nice words in good-looking update
> > announcements, which advertise everything but the regression and
> > breakage. How does that help us? The testing is missing, not the
> > announcements.
Exactly - marketing eye-candy.

> Correct the testing, 
Wrong, correct the release procedures. E.g. to me it's incomprehensible
why Fedora doesn't automatically block out packages with known broken
deps.

> forcing things to go into an updates-testing queue before 
> going out to the rest of the world.
Once, again, this only makes sense if anybody uses it and if those using
it are able to perform real and systematic tests.

Experience tells, such "testing" is largely being ignored by the masses.

Experience tells, packages in such "testing" repos are only used by
users if they are facing real problems (such as with the kernels).

Experience tells, hardly anybody is able to perform "real and systematic
tests" and if maintainers are able to handle reports.

Experience tells, maintainers often ignore negative feedback on
"tests" ("Your case is a rare corner-case, I won't update the package
for it") or can't avoid to fall back to "sorry, I can't do much about
it".

>   Announcement about what has changed and 
> what should be tested, the ability to pull something from -testing before it 
> hits the wide world.  Ralf's statement is that we shouldn't bother, 
Announcements don't change anything about it.

> and we should just use our end users for our testing purposes.
I say: Just as throughout all the years before, you should use rawhide
for testing and avoid any disruptive changes on stable.

> > Anyway, I wonder what this has to do with a sudden and late freeze of
> > Extras' packages?
Not much, except that forcing maintainers to manually writing
announcements (which will widely be ignored) is yet another additional
regression in workflow imposing further load on maintainers.

Ralf