Schedule figures

Jon Stanley jonstanley at gmail.com
Tue Sep 23 23:30:17 UTC 2008


On Tue, Sep 23, 2008 at 5:04 PM, Paul W. Frields <stickster at gmail.com> wrote:

> Probably, and they need to come from people smarter and more experienced
> than me.  We might want to involve James Laska in this thread as well.
> By the way, I don't know who-all's here on the rel-eng alias, but I am
> assuming that this is an open-sub list -- I'm sure I've visited the
> archives before so I know they're open.  If I'm wrong, someone tell me
> so we can get this into the right open venue please.

This is the correct open venue, though admittedly one that probably
not many people know about :). As far as the problem at hand, freezes
are meant to be just that IMO, however, events warrant getting things
in after the freeze nearly all the time, for example things not
installing right :)

I firmly advocate a strict freeze policy, basically something like "we
only accept items after the freeze if the default installation lights
itself on fire, or you're unable to do an installation of various
package sets via various methods", with the various methods becoming
more and more towards the END of the release process.

For Beta/Feature freeze, the situation becomes a little tougher.
According to the feature policy, any features for the current release
have to be in a testable state by Beta. This leads to people
requesting feature-related tagging. I'm pretty sure that I'm being
very unreasonable here, but a side of me says "if you don't have your
feature in a testable state by Beta *freeze* - we drop the feature".
No more tagging just to ensure that features meet the policy (again,
I'm very sure that this is an unreasonable position to take - but it
does represent the ideal world).

This sounds counter-intuitive, I know - but we expect less to work
towards the beginning. Towards the end, there are more stringent
requirements that we are ready for the world. Admittedly, in a perfect
world, everything would be tested for every milestone release that we
have, however, we have a distinct lack of QA resources in Fedora -
Will is just one person, and while he does have help from others, the
time that most of them (including myself) spend is at the moment that
we have time available, not necessarily at the time of need :/


More information about the rel-eng mailing list