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

[Ralf Corsepius]

On Tue, 2007-05-15 at 13:25 -0400, Will Woods wrote:
> On Tue, 2007-05-15 at 18:31 +0200, Ralf Corsepius wrote:
> > On Tue, 2007-05-15 at 12:11 -0400, Will Woods wrote:
> > > On Tue, 2007-05-15 at 17:00 +0200, Patrice Dumas wrote:
> > > > Will there be a way to short-circuit the announcement and just update,
> > > > like it was possible before?
> > > 
> > > Holy Zod, no. No no no. Very no. 
> > > 
> > > There will be no more silent updating of packages without explanation
> > > or at least a *chance* for testing. 
> > What you call testing, I call locking out updates and bug-fixes
> > == regression
> 
> Excuse me? Who said anything about locking out updates or bug fixes?
It's what already is happening due to this new freaking release
procedure.

>  I'm
> just asking that packages spend a few days in a -testing repository -
> which is open to the entire world - to be sanity tested by QA and
> interested users before going to the updates repo.
You are not able to test 1000s of packages nor are you able to have
tests for them. All what you might be able to do is performing some "rpm
consistency checks", but that's all.

> How does that lock out *anything*? What are you even talking about?
* E.g. about API, SONAME changes, compat-packages etc. If you're going
to freeze APIs, SONAMEs etc. you're locking out packages.
* By forcing maintainers to fill out forms or similar steps, you are
imposing additional load on maintainers. I for one don't have any time
available to waste on this kind of kid stuff. As a consequence you'll be
facing me not upgrading packages.

> Where are you drawing these ridiculous conclusions from?
You are telling me you are "testing" - I say, _you_ can't test, because
this is technically impossible.

> > > Yeah, I know this makes putting out updates slightly harder,
> > 
> > == regression
> 
> You keep using that word. I do not think it means what you think it
> does. 
No, I really mean 'regression": You are breaking and complicating what
has been functional and simple before 
=>  "technically harder == usability regression".

> Let me get this straight: You think that being able to cram changes down
> the throats of ~2mil users without justification or testing is a
> *feature*, and that requiring (world-accessible) testing is a *bug*. 
Yes. It's a waste of time, because 

* a "responsible/conscious maintainer" is supposed to test his stuff in
advance and should be better knowledgeable about a package's details
than anybody else. If he isn't, he probably should not maintain this
package.

* a mandatory "update-testing" is superfluous in the vast majority of
cases, an optional "update-testing" can make sense in _some_ (few)
cases.

* a mandatory "update-testing" repo is of very limited use to the
majority of users, because only those facing issues should use it for
"selected updates". Ordinary users should not use it all. Un-educated
users however will activate it.

> > >  but in my
> > > opinion the fact that we ever allowed this at all was a massive
> > > oversight, not a design feature.
> > 
> > I disagree. Release early, release often is a feature, now you are
> > spoiling one of the fundamental working principles of Fedora.
> 
> How is it spoiling anything? 
Delays, bloat, complexity, bureaucracy, featuritis.

> The packages are still available
> immediately after build, just like before. I think you have
> misunderstood what we're talking about here.
Possibly, because RH once again seems to have failed to communicate what
THEIR plan is and seems to be pressing something which doesn't seem to
be clear to themselves onto the community.

At the moment I am primarily referring to this nonsensical regressions
this new release flow and koji impose on my packages. I see regressions
all over the place: What once was simple, now requires additional effort
and wastes my time.

Ralf