Re: Better collaboration with the GNOME project]

[Frederic Muller <fredm gnome org>]

One more response from Canonical.

I'll synthesize current response on the GNOME wiki and provide a link after.

Thanks.
Fred

-------- Original Message --------
Subject: 	Re: [Fwd: Fwd: Better collaboration with the GNOME project]
Date: 	Wed, 23 Feb 2011 08:22:13 +1030
From: 	Jason Warner xxxxxxxxxx
To: 	Frederic Muller



Hi Frederic -

[...]

Here are the responses to your questions. It might be worth a
conversation over a beer sometime. Perhaps at Desktop Summit?
BTW...several members of the desktop team answered the questions below;
Martin Pitt and Sebastien Bacher. I'm more or less sending it along and
adding some minor commentary.

========================================================

Some of the initial information we'd like to gather are related to long
term support of specific GNOME releases:

1. What version(s) of GNOME do you maintain in a stable fashion?


    8.04 LTS:   2.22.2
    9.10:          2.28.1
    10.04 LTS: 2.30.2
    10.10:        2.32
    Natty:         2.32.x


2. How much work does this represent?

For our released Ubuntu versions it takes an order of two manweeks to
update to new upstream microreleases after releases, and maybe another
one or two manweeks in total to sort out major regression bugs.

For the development release it's a significant and permanent effort,
as we not only package new releases (which is already a major
undertaking by itself), but also help with bug and integration fixes,
develop some features of our own, etc. For the actual GNOME parts (i.
e. not counting indicators, Unity, etc.) I'd say it's the equivalent
of two full time jobs (it's actually done by a lot more people,
including community, of course).

3. Do you feel there is duplication of work between what you do and what
other distribution do?

Yes, there is:

  * Packaging in a strategic sense; of course Fedora/SUSE RPM packaging
   is totally different from Debian style packaging, so there is only
   little potential for reducing the duplication here.

   This is a lot better with Debian, of course, as we regularly merge
   and have the same packaging format and policy.

  * Bug triaging; each distro has their own bug tracker, which provides
   the "front line" for users to report problems, distros to sort
   them, weed out duplicates, analyze them properly, and then submit
   good bug reports upstream. There is certainly a lot of duplicated
   bug triaging going on here.

  * The parts of distros which do software development work on
   different projects for the same purpose. Standard example here is
   Unity vs. GNOME Shell.

4. How do you see a potential collaboration between all of "us"
(upstream and downstream projects)?

* It is vitally important that all non-distro specific changes,
   fixes, and new features get submitted upstream, and usually
   developed against the current upstream git head. This is not just a
   potential collaboration, but actually happens. We (and everyone)
   can always get better, of course, but it would hurt us a lot and
   bind way too many resources if we would not do it.

  * Over time, many folks in the desktop team also have become upstream
   committers, so in some cases we can develop/fix stuff directly
   upstream and avoid distro patching entirely. IMHO this is the ideal
   state which distro developers should aspire to reach for their key
   packages.

5. We are definitely aware that today each of us use a different bug
tracking system. Do you see any possible technical solution that could
address this specific issue?

It is not so much a technical, but a social and cultural
problem. So far we actually deliberately keep the distro bug trackers
separate from upstream's. The average Launchpad bug is of rather low
quality, and often need some developer brain power to turn it into a
bug report which is suitable for upstream, and the majority of
bugs aren't suitable for upstream forwarding at all (duplicates, not
reproducible, using software in a wrong way, etc.).

In theory we could merge all distro GNOME package bug trackers into
the single upstream bugzilla, and everyone works from there.
Personally I would not do this, though, as it would basically destroy
the usefulness of the only bug tracker (upstream bugzilla) which is
still a reasonably effective tool for actual developers.

But in the end there is no golden bullet here; bug reporting/triaging
is inherently very prone to redundancy and requires a lot of detective
work, as long as we allow everybody out there to report bugs.


-----