To freeze, or not to freeze

Jon Ciesla limb at jcomserv.net
Mon Sep 8 16:38:03 UTC 2008


> On Mon, 2008-09-08 at 11:31 -0500, Jon Ciesla wrote:
>> > On Mon, 2008-09-08 at 10:27 -0500, Jon Ciesla wrote:
>> >> > A few weeks back the releng team made a schedule adjustment, to
>> make
>> >> up
>> >> > for lost rawhide time due to the intrusion.  When we made that
>> >> schedule
>> >> > adjustment, we had assumed having a working rawhide within days of
>> >> that
>> >> > meeting, giving us a week or so worth of working rawhide to shake
>> out
>> >> > any pre-beta bugs we wanted to.
>> >> >
>> >> > Unfortunately, while we were able to /attempt/ rawhide creation as
>> >> > scheduled, a series of software bugs and bad personnel timing has
>> >> > prevented those attempts from producing usable rawhide images until
>> >> > basically today (and today's are pretty shaky from what I hear).
>> This
>> >> > is not so good, as without an installable rawhide, we don't get a
>> good
>> >> > idea as to how the installer is working, and we miss out on a lot
>> of
>> >> > 'initial install' testing of software on people's systems.  We get
>> a
>> >> lot
>> >> > of "I upgraded from foo" type testing, which definitely has it's
>> >> value,
>> >> > but I feel we're missing a pretty key part of the rawhide
>> experience.
>> >> >
>> >> > The Beta freeze is set for Tomorrow.  That means the content that
>> >> would
>> >> > show up in tomorrow's rawhide would also be the content we use as
>> the
>> >> > basis of Beta.  Any changes after that would have to be ran through
>> >> the
>> >> > releng/qa teams to be approved.
>> >> >
>> >> > Given the "fun" we had with Alpha, I really feel that it would be
>> >> > prudent to spend a few more unfrozen days with a hopefully
>> continually
>> >> > working rawhide installer so that we can do some of that last
>> minute
>> >> > testing of what's wrong before we freeze, and hopefully have a
>> shorter
>> >> > and more productive freeze period.  But this is just my opinion,
>> and
>> >> > thus I'm putting this out there for discussion.
>> >> >
>> >> > Feature owners in particular, I'm interested in your opinions as to
>> if
>> >> > you need a few more days to see what shape your features are in
>> before
>> >> > we freeze.  Ideally the week before a freeze would have been a slow
>> >> down
>> >> > period, where large changes were avoided and bugfixing was focused
>> on
>> >> so
>> >> > that the Beta was useful.  I feel like we didn't give you a chance
>> to
>> >> do
>> >> > this and it'll still feel like crash landing planes on the carrier
>> >> deck
>> >> > when it come to features in Beta.
>> >> >
>> >> > I'm interested in what the rest of you think as well, both package
>> >> > owners and testers alike.  If you think our schedule time would be
>> >> > better spent fixing and verifying things pre-freeze and adding an
>> >> extra
>> >> > week to the schedule, or just freeze as things are, and potentially
>> >> slip
>> >> > a week during the freeze to make everything usable for the Beta (or
>> >> the
>> >> > third option, things are fine as they are, just freeze and release
>> as
>> >> > scheduled and stop being so paranoid).
>> >> >
>> >> > What say you?
>> >>
>> >> Wait a bit, maybe slip the release a week if you have to.  Better
>> >> slippage
>> >> than suckage, IMHO.  After the intrusion, I think users will
>> understand.
>> >
>> > slipping the release a week gets complicated as we get dangerously
>> close
>> > to thanksgiving in the US which takes a huge number of contributors
>> out
>> > of the world.
>>
>> Are you suggesting that after a large meal containing high quantities of
>> starch, tryptophan, and alcohol, I might not be at my best?
>>
>> For shame. ;)
>
> No, I'm saying most of us don't want to work through that weekend.
>
> I know I don't. :)

Agreed and seconded.  <Sigh>  Would have made for some memorable cvs
commit messages, though.

I think we could easily make up the time between F10 and F11 without
slipping F11, though.  Especially if there isn't another intrusion.  We
just might have a really shaky alpha.  So what?  It's an alpha.

> -sv
>
>


-- 
novus ordo absurdum




More information about the fedora-devel-list mailing list