On Mon, May 19, 2003 at 03:22:11PM -0400, Chris wrote:
> Let me begin by stating that this is only an attempt at constructive
> criticism and an offer of help. I do not mean to nag.
Your message didn't come off as nagging at all. Thanks for sending it.
> What frustrates me personally is not the fact that official 4.3.0 packages do
> not exist but rather that there is nothing available for experimentation.
That's not literally true; Daniel Stone's packages have been available
for weeks (months?) and have generally been received well.
deb http://www.penguinppc.org/~daniels/sid/$(ARCH) ./
deb-src http://www.penguinppc.org/~daniels/sid/source ./
> I am fully willing to help as I can, but I see no avenues to do so.
Hopefully this message will provide some pointers, and condense some
information that I've been sharing on a sort of ad hoc basis with
people.
> For brevity, my suggestion is this: openness, delegation and distribution of
> work load.
I couldn't agree more. Sounds like my platform for Project Leader. :)
> If the process of packaging xfree86 for Debian is too much for one
> person to do in their spare time with reasonable timeliness, I believe
> it may make sense for those capable to collaborate with Branden to
> lighten his load in whatever ways they can.
I've put out calls for assistance several times over the years, most
recently on April 16th[1], but these have seldom borne any fruit.
This mailing list has accrued several highly clued people who are
competent to address questions that I am not (particularly diagnosing
various forms of X server breakage, especially on hardware I don't have
access to), and I feel the load is being more broadly shouldered than it
used to be.
However, as you note, much more remains to be done.
Openness
--------
On the openness front, I believe I have already made a great stride
forward. Last month I set up a Subversion repository to house my
development efforts on several packages, including XFree86, Xft, and
Xrender. However, I have not publicized the existence of this resource
as widely as I might because I was nervous about ViewCVS's
(experimental) support for Subversion repositories rendering the repo
inaccessible (this has actually happened several times). However, I
haven't seen this happen in a couple of weeks now, so your message
serves as a good opportunity to give it a little more publicity.
http://necrotic.deadbeast.net/cgi-bin/viewcvs.cgi/
Delegation
----------
The first step has already been taken on this front. I invited Daniel
Stone to use my Subversion repository for his XFree86 work, and he
immediately and enthusiastically did so. I created a restricted
account for him on the above server that enables him to use Subversion
over SSH to update the XFree86 repository arbitrarily.
I hereby extend an invitation to other people on this list to send me an
email if you'd like to use my XFree86 Subversion repository. Please
familiarize yourself with Subversion (if necessary) and my repository's
layout[2] first. Just let me know what you'd like to work on. If
you're a known quantity in the Debian Project such that I don't have to
worry about you vandalizing the repo, I'll be happy to grant you access.
One thing that should be noted is that only the "debian/" directory of
the XFree86 source package is currently being managed in my Subversion.
Subversion's support for tagging and branching, while quite powerful,
imposes unacceptable disk costs in checkouts for a project of XFree86's
size, especially if one imports all of the CVS tags and branches from
the XFree86 CVS repository.
I think XFree86 could benefit tremendously from a team-maintainership
approach, as Debian's packages of the GNU C Library, GNU Compiler
Collection, and others already do. As my role and responsibilities in
the Debian Project and for Software in the Public Interest, Inc., have
grown over the years, I have found myself less able to devote
single-minded attention to XFree86 itself. I completely agree with you
that this needs to change. While I think I can bring more value to the
Project as something more than just the XFree86 package maintainer, the
packages need as much attention as I was able to give them at my best.
Probably more.
Distribution of Workload
------------------------
This pretty much dovetails with the previous point. Before developing a
comprehensive strategy in this email message, I'd like to see how much
response I get to my solicitation for co-developers in my Subversion
repository.
That said, here are a few tasks that are readily distributable:
* modularization of the XFree86 source tree, as recently discussed on
the XFree86 forum mailing list; there is no particular reason that
buffer overflow fixes et al. in Xlib should force a simultaneous
release of the XFree86 X server
* aggressive and systematic triage of bug reports, especially old
upstream bugs; XFree86 upstream as of this past March has a Bugzilla
system in operation, so it now may be meaningful to "forward" bugs
upstream
* specialization in certain areas of the upstream codebase, perhaps
selectively patching from upstream CVS HEAD or independently-developed
patches
So, how about it folks? Shall we make XFree86 a little-less of a
Branden-monopoly?
[1] From: Branden Robinson <branden@debian.org>
To: debian-x@lists.debian.org
Subject: Re: Maintainers with excessive old RC bugs
Date: Wed, 16 Apr 2003 14:12:26 -0500
Message-ID: <20030416191226.GR4365@deadbeast.net>
[2] http://necrotic.deadbeast.net/cgi-bin/viewcvs.cgi/README?rev=8&content-type=text/vnd.viewcvs-markup
--
G. Branden Robinson |
Debian GNU/Linux | // // // / /
branden@debian.org | EI 'AANIIGOO 'AHOOT'E
http://people.debian.org/~branden/ |
Attachment:
pgpkOmUfQs3Tt.pgp
Description: PGP signature