Fedora 11 schedule proposal

[Jeremy Katz]

On Tue, 2008-11-11 at 16:11 -0800, Jesse Keating wrote:
> Fedora releases typically have a 6 month development cycle.  We target
> specific dates for the release to give developers, end users, and
> upstreams a target to shoot for.  Typically any slipping of a release we
> do, we just shorten the next release to make up for it.  However a
> month's time is quite a lot to shrink.  Especially because of the
> significance of F11.

FWIW, the past slippage of a month that we had, we made up the month
over the course of 2 release cycles to help reduce the impact to each
individual release.  

> Fedora 11 will be extremely important to Red Hat Enterprise Linux
> (otherwise known as RHEL).  RHEL 6 planning has looked to use Fedora 10
> and Fedora 11 as releases to work out new technologies and features that
> are desired in RHEL 6.  This includes a lot of upstream work that is
> being done, and targeted to land in these two releases.  That planning
> was also planning for a full 6 month Fedora 11 cycle, and Red Hat
> resources were lined up to take advantage of this, by participating more
> in the development cycle, in the testing cycle, in bugfixing, etc...
> This is a good thing.
> 
> However, if we were to take a month out of Fedora 11s schedule to hit
> that May 1 date, we would shorten the amount of time we get the RH
> attention, and we shorten the amount of time we give our developers to
> land the pre-planned features.  This is not a good thing.  These are not
> just RH developers for RH features either, it's all developers for all
> features.

So, I don't fully buy this reasoning.  As you said above, we target
consistent dates for each release.  This is to help developers (upstream
and downstream) know when they need to target having things done.  And
given that we try to do most of the work we have in Fedora in upstream
projects as opposed to in a Fedora silo, a slippage of a Fedora release
fundamentally doesn't change when things would need to be upstream.  So
I don't see how the fact that we slipped our release due to
infrastructure problems shortens the amount of time developers have.
They had until May 1st before, they still have until May 1st (well,
before that due to freezes; but you get the idea :)

> But wait, if we pad Fedora 11 by a month, what about Fedora 12?
> Excellent question.  We (releng) feel that if we set up Fedora 11 with a
> full 6 month schedule, and now set up F12 to take in some of that slip
> and shoot for Oct 31 release of F12, all interested parties will have a
> lot more time to prepare for a short release.  In fact, we could even
> focus more on polish issues in F12 than large sweeping features, but I
> leave that up to the developers.  We just feel that we can better plan
> for a short release in the F12 time frame than we can in the F11 time
> frame, and by doing so we can get back to our May 1, Oct 31 established
> dates which work pretty well in the Linux ecosystem.  Of course, setting
> an F12 schedule now would be kind of silly given that we don't know what
> will happen during F11 (hey, look at what happened to F10!), but we can
> at least have an idea of what we'd like to do in F12.

How does more time help us to prepare for a shorter release?  If more
time helps us prepare, why weren't people preparing already for this
with Fedora 11?  And as was alluded to by bpepple, we actually have a
lot less freedom with a short fall/winter release due to the proximity
of the holidays (which we're currently wrestling with)

Also, I suspect that this problem will be even worse for a shortened
Fedora 12 where many Red Hat resources are concentrating on RHEL 6.
It's been a while since RHEL5, but the inability to get people's
attention on things for Fedora 7 was one of the (admittedly many)
reasons that it actually slipped out from the six month schedule.

If we're going to do a shortened release, we should make it be the one
that has the most possible attention from all stakeholders :-)

> Long story short, here is a simple schedule we'd like to present for
> F11.  http://fedoraproject.org/wiki/TomCallaway/F11ProposedSchedule

I'd personally prefer that we keep to the dates that we have said and
are expected by the rest of the world.  Fedora 11 has its target date of
the nearest Tuesday to May 1st (May 5th if I'm reading my calendar
correctly).

If, instead, we strongly feel that we need a longer schedule for Fedora
11, let's look at doing it for real.  What actual features are people
looking to land for Fedora 11 that won't be ready in time for a May 1st
release and the corresponding beginning of March or thereabouts feature
freeze?  From that, we can come up with what the schedule should be to
accommodate those features and we accept that Fedora 11 is a
feature-driven release with its schedule set based on that, rather than
being calendar driven.  And if we do this, we shouldn't do it in a
vacuum ignoring Fedora 12's schedule -- instead, we should look at what
Fedora 12 looks like and take that into account.  Maybe that over the
next 18 months we only have 2 releases as opposed to 3.  But then we're
consciously making that decision based on what we're trying to
accomplish with the releases rather than just trying to make schedules
look pretty on paper.

Jeremy