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

Re: PROPOSAL: preparation for freezes, release coordination

On Sun, May 09, 1999 at 03:43:58PM -0400, Adam Di Carlo wrote:
> Looking back at the slink freeze, I think we had two big problems
> which slowed us down by about a month: new X Window System packages,
> heavily broken, and the libc problems.  I don't wish to cast any blame
> on the package managers for this -- I just hope we don't have a
> repeat, and I hope everyone learned a lesson from that.

Well, the problems with X were obviously my fault.

I got caught between a rock and a hard place -- I had promised that I would
reorganize the binary packages for the next release (which was slink, at
the time), and made that promise literally months before slink froze.  In
fact, I think I had made up my mind by the time we released hamm that it
was something that needed to be done.

I made that promise without realizing just how *EVIL* the whole things was.
God, X is complicated; maybe I am too simple-minded for it, but I really
think you can't get a grasp on just how much of a monster it is until you
*really* try to wrestle with it.  Grabbing the sources, fooling around in a
few .cf and .def files and typing "make World" just doesn't come close.

I apologize for being the cause of substantial additional delay in
releasing slink.  I wasn't very loud with my penitence before because it
was a sore spot with me.  Now that I even have out, people nag me a
lot less and I feel better.

Nevertheless, I will defend the actual act of the Great X Reorganization to
the death.  I am convinced that X is much more logically packaged now;
that, coupled with the "second Great X Reorganization", which is already
complete, will make this package much easier to maintain.

The binary package reorganization grouped X into much more logical units,
consistent with other Debian packages.  It also modularized things to a
much greater degree so that people didn't have to install things like twm
and xterm if they weren't going to use them.  I think this was a big plus.
It also makes managing the bug list easier because 75% of all bugs aren't
one package (xbase) anymore.

The source reorganization, which was just implemented (largely thanks to
Adam Heath) for potato, will make packaging of future new upstream releases
much easier.  That's the other big complaint people have about X.

So, this time, I've got my big ugly changes (which in this instance aren't
that visibile to the user, because the binary packages have hardly
changed) out of the way.

In exchange for being a good boy this time, I'd like some time to implement
some important changes.  Among the things on my "Looking ahead" list at the
X Strike Force page, a few things spring to mind as especially important:

1) See what we can do about getting rid of xterm's setuid bit.  Wichert
just made a proposal that may solve this.
2) Get the Unix98 PTY stuff firmly integrated, and get xterm writing to the
utmp, wtmp, and lastlog files all appropriately.
3) Revisit the much-loathed "xterm-debian" issue and see if I can come up
with something that will make people in heterogeneous environments happier.
(If the X Consortium hadn't released a ballsed-up xterm in X11R6 all those
years ago, much of this frustration wouldn't have happened.)
4) Get xfs running as its own user, not as root.
5) Further examine PAM-ification issues of xdm and the X server.  The X
server might be straightforward, but apparently there are deficiencies in
the design of xdm as far as good PAM support is concerned.

G. Branden Robinson              |   I must despise the world which does not
Debian GNU/Linux                 |   know that music is a higher revelation
branden@ecn.purdue.edu           |   than all wisdom and philosophy.
cartoon.ecn.purdue.edu/~branden/ |   -- Ludwig van Beethoven

Attachment: pgpaS0EbgYO7S.pgp
Description: PGP signature

Reply to: