[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Xorg 1.5 missed the train?



On Sun, 2008-05-11 at 11:37 +0300, George Billios wrote:

> > Upstream maintainer and people
> > closely following development think of a module as a list of new
> > features and current bugs, shrinking and growing by time.  If Adam
> > thinks it's fine to include a prerelease of xorg in Fedora, then, well,
> > it's fine.
> 
> You (and others here) miss the point, that not everybody is a developer, 
> not everybody monitors the bug list but almost everybody can notice the 
> word 'prerelease' in the F9 release notes and would like a reason why 
> this happened.
> 
> Take this for example:
> 
> http://fedoraproject.org/wiki/Features/XserverOnePointFive
> 
> It mentions that it is a prelease version but it doesn't justify why 
> including a prelease version is ok.

It's the least buggy X server branch with the features we want.

I admit, it's not a release, and that's entirely process failure on my
part.  Having lots of masters to obey is not easy, and in this case my
time got chewed up by other business obligations.  Thanks RHEL, you're
awesome.  So the thing I chose to sacrifice was the (actually fairly
labor-intensive) process of badging the tarball as a release.  It still
got bug fixes.  It's ABI-stable.  Leaving it in was way less disruptive
than reverting back to 1.3 would have been.  It just isn't 1.5.0.

I actually went on a long rant about this at xdevconf:

http://xorg.freedesktop.org/wiki/Events/XDC2008/Notes

The essential problem is that we're a victim of our own success.
There's lots of great stuff happening in X and unfortunately the thing
to suffer there is the stabilisation effort.  I do the best I can but
I'm not a superhero.  Nor, really, should I have to be.  And come to
that, I'm actually quite bad at the release management job.  I don't
communicate effectively, I don't delegate enough, and I don't deliver on
the schedules I promise.  Mea maxima freakin' culpa.

If someone else steps up to the job, huzzah.  Until then we're sort of
stuck.

My question to the gallery is how do I fix this?  How do _we_ fix this?
X needs more people.  It's way less scary than you've been led to
believe.  How do we get more people involved?  How do we step up the
testing effort?  How do we get to a culture of frequent releases and
incremental improvement?  Without these things, release management is
going to continue to be bursts of heroic effort that almost certainly
misses deadlines, as it has been ever since Xorg 6.7.

I mean, in some sense, it's fine.  It's software just like any other,
the number on the side of the box is merely a talisman, a dusting of
holy penguin pee.  But the release is also the primary artifact of the
development process.  Skipping that obligation is a disservice both to
ourselves and our consumers.

I hope we find an answer, but I just don't have one right now.

- ajax

Attachment: signature.asc
Description: This is a digitally signed message part


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]