What next?

Michael Tiemann tiemann at redhat.com
Sun Jun 5 23:07:21 UTC 2005


On Sun, 2005-06-05 at 12:30 -0400, seth vidal wrote:
> On Sun, 2005-06-05 at 04:27 -0400, Daniel Veillard wrote:
> > On Sun, Jun 05, 2005 at 03:08:40AM -0400, seth vidal wrote:
> > > 
> > > > Projects are getting larger and larger and need more time to develop.
> > > > Increasing the release cycle every year or two isn't the answer.
> > > 
> > > How would you know? Red hat has never had a release cycle of anything
> > > other than 6 months. You seem to be making an assertion you can't
> > > support. I'm suggesting we try something to see if it helps.
> > 
> >   Sure we do internally... RHEL so far has had an 18 months cycle
> 
> Nice of y'all to allow yourself longer release cycles while denying it
> to fedora community developers.

Not fair--see below.

> > and it does work by branching at some point and doing the work we want
> > to do on the community base until it is ready to release. Like you we
> > need more time to integrate changes and stabilize than what the 6month
> > cycle allows, we also have the need to integrate most of it back into
> > the main cycle, and not disrupt it. It requires planning, branching,
> > merging, but it's doable and I think we can support that model from a
> > technical standpoint since we are over the fourth iteration on that model.
> 
> I think we don't have nearly enough volunteers or infrastructure to say
> that the fourth iteration of that model is viable. I've watched the
> amount of stuff that needs to be done at the fedora extras steering
> committee meetings and it's non-trivial and needs to be done ASAP.

Seth, you've actually answered your own question.  Based on what I have
seen inside of Red Hat, 6 months is a release cycle that matches well
the challenge of the problem with the nature and resources of the
present community.  Even with the surprise 7 day delay, FC4 is great, as
was FC3 vs. the targets we all set.  I even think that your suggestion
of a one-time-for-now 9 month cycle could make sense.  But the stuff
that's required to hit an 18 month target is just Night And Day
different.  It may look easy--copying and criticizing the decisions we
make is pretty easy stuff compared to sorting out the initial plan and
keeping enough of the contingencies viable that the release is relevant
by the time it sees daylight.  I'm not saying that Red Hat is the only
one trying to make these guesses and trying to following them through to
their logical conclusion--there are companies and community efforts
alike that try this stuff every day.  But I am saying that it's a
sufficiently difficult and resource-intensive task that it's Just Not
Fair to suggest that simple goal-setting will lead to simple goal
attainment.  Running a marathon is hard, but running up up Mt. Everest
is a different ballgame, and pretending it's not is reckless.

My $0.02.

M




More information about the fedora-devel-list mailing list