Lessons Learned

[Michael Schwendt]

On Wed, 21 Mar 2007 07:20:22 +0100, Thorsten Leemhuis wrote:

> Hi!
> 
> Josh said some good stuff already, no need to repeat all of it.

True.

> FESCo was good (but still far from perfect) in getting the community
> involved, as it had influence in certain decisions and meets in the
> open.

Different than that. It had started as a group with the desire to move
several things forward at the Extras front *actively*, which included
boring but necessary tasks. 

Red Hat in turn made a few allowances for FESCo in order to remove some
road blocks. A necessity. Else Extras would not have taken off. I wouldn't
call the result "influence", though. The influence is mostly by a very few
individuals, who sponsor hardware and access to it or who make things
happen inside Red Hat.

> With the merge happening now it seem FESCo seems to me much less
> important then before; to much small to medium stuff gets simply done
> without even bothering to ask FESCo -- that's a fault IMHO. A lot of
> medium to big stuff gets directly taken to the board -- FESCo afaics
> often gets circumvented, even if the tasks are engeneering tasks. And
> that's what FESCo is for afaics.

FESCo fails to define its field of activity. The moved Wiki pages for the
new FESCo don't add any description, but are still in CategoryExtras, and
it seems FESCo just picks random things that pop up here and there. There
is no hint as whether any group still leads Extras -- trying to remove the
word "Extras" everywhere is not the full show.

What I'm missing is the regular presence of FESCo or its members in
official day-to-day decisions and state-of-the-union addresses, so the
community of contributors gets the feeling that there is some kind of
leadership actually. Instead, it has increasingly become a sit-and-wait
process, where hardly anyone seems to care about some things until
somebody else complains.

It also takes too long to create a low-traffic announce list, where to
reach package maintainers actually. fedora-maintainers apparently is not
the place where to reach them. Cross-posts to at least -devel, -extras and
-maintainers are way too popular. Annoying. Surely the reason is that
hardly anyone knows what the purpose of those lists is. Well, I don't
either, looking at the thread index.

When I remember how often we've practised preparing Extras for the next
release of Core, I believe this time we perform badly. We have Matt
Domsch's rebuild reports, the broken deps report, the broken upgrade paths
report, a FE7Target tracker bug filled with lots of bugs. Are any of the
reports on the mailing-lists evaluated by anybody in FESCo? It's
interesting, for instance, that Zope and Plone, which are in the broken
upgrade paths report for a very long time, don't support Python 2.5. Yet
Extras bugzilla did not cover that. Huh? So:

  https://bugzilla.redhat.com/233187
  https://bugzilla.redhat.com/233185

Nearly all the other upgrade path problems were not in bugzilla when I had
a look yesterday. A big bunch now is in the FE7Target tracker, at least.
Several I've fixed myself, but ACLs block access to other packages. I'm
disappointed that FESCo (with 'E' as in Extras) in no longer involved in
trying to reduce crap in the distribution, not even with a roll call.
I don't mind the AWOL process, which I've tried with one maintainer I'm
unable to reach for several weeks. But we're at test3. Time is getting
short if we still want to offer something that doesn't bear too much of a
risk of confronting users with broken deps and non-working components --
especially after Eric Raymonds complaints about [albeit house-made]
dependency problems in Fedora-land.