Re: glibc 2.1 and compatibility (Was: slink is gone, goals for potato?
On Fri, Mar 05, 1999 at 09:04:23PM -0700, Steve Bowman wrote:
> What I have in mind is to divide up the packages into groups to make
> the freezing process more "robust" (if you can apply that to a process).
(Oh heck, I'll let myself get drawn into this discussion. This isn't all in
response to your post.)
I believe that we should address this problem in the Unix tradition of
prefering simple abstractions to complicated ones, and the free software
tradition of prefering cooperation and communication to formalized
infrastructure.
The "develop, freeze, release" paradigm has proved itself--it is responsible
for Debian's reputation for superior quality, and should not be altered
lightly. So, before proposing a more detailed model, let's see how we can
get more mileage out of the existing one. In particular, how can we address
the current "Debian is obsolete" criticism?
- One obvious and much discussed step is to shorten the release cycle and
(particularly) the freeze. While suggestions like "freeze potato two
weeks after the slink release" may provide a small gain, a better approach
is simply to improve communication among developers so that everyone knows
where the distribution stands with respect to the next release, and how
they can best contribute. This was evidently lacking in slink--for
example, the dpkg bug that delayed the release "at the last moment"
appeared unheeded on many editions of the release-critical bug list.
Richard Braakman (sp) has described how he would address this issue, and I
believe he would be a fine release manager.
- Another idea is to make newer packages available to users of stable
releases. Here we seem to be stuck between maintaining an additional
track of development or making stable unstable. I proposed a cheap way
out: compile certain packages from unstable on stable releases, and
provide them, without warranty, in a separate distribution clearly labeled
as untested. This would be good enough for most complainers without
causing Debian much extra work.
A distribution that's both up-to-date and well-tested is a nice dream, and
it is tempting to think that with a more complicated system Debian could
achieve it. But it's much more practical to think about how we can do more
with the current system.
Andrew
--
In general NT is what you get if you could do some portions of Unix all over
again, and then managed to get it all terribly wrong.
- Maury Markowitz <maury@remove_this.istar.ca>
Reply to: