(time to do some apologizing)
> This really annoys me. (No, not you Jason.) I agree that we have to
> move on the the next distribution, but given the fact that
> <speculation> *most* </speculation> debian 1.3.1 users use netscape
> and that <speculation> *most* </speculation> of those folks will want
> to use a plug-in, I think this bug merits an upgrade to 2.01 bash
> for 1.3.1? If a really .deb upgrade is not the answer, can we create
> a special upgrade on ftp.debian.org where I can upload a 2.01 bash
> that I've built for 1.3.1?
Does bash 2.01 solve the problem? We do update 'stable' - we're
currently debating that strategy on the debian-private (developers only)
mailing list right now. If bash 2.0 is sufficiently broken, then that might
merit putting 2.01 into 'stable'.
> In fact while I'm at it let me expand on this general point. Let me
> say first that I marvel at the (apparent?) organization inherent in
> the debian development "system". Even though there are many maintainers
> everyone seems to be on the same page. However, I disagree with the
> philosophy of fixing bugs in "old" distributions. Who decides when a
> bug is important enough to be rolled back into an old distribution.
Basically the maintainer of the package does (so take up the issue
with him). I'm basically opposed to upgrading the 'stable' distribution
after it's been released -- except for packages that have security
flaws or were released in a 'broken' state (we shouldn't be doing that).
If someone really needs the bash 2.01 version - they can always install
it out of 'unstable' (along with libc6). No problem.
> I use debian at home and at work. For home use, I like to be on the
> cutting edge. At work, my job is developing software for
> clients, not being a systems administrator. I want a distribution
> which is easy to maintain and as well thought out as debian. Currently,
> soon after a distribution has been released (this happened for 1.2
> as well as 1.3) it is abandoned by the developers. There are *still*
> bugs in the 1.3 installation (such as the problems with X). And why
> the hell was X 3.3 tossed in at the last minute? The 3.2 dist. didn't
> even install right.
<excuses>
Well, a whole bunch of security bugs popped up at the last minute.
XFree is one of the biggest packages, so the X maintainer was swamped.
It was a lot less work to package up the just-released 3.3 version.
Unfortunately, all of this happened just as Debian 1.3 was being
released. There was no opportunity to thoroughly test it in unstable.
</excuses>
The other major problem is that we are allowing broken packages into
'stable'. There was also a problem where nobody realized that some of
the X install bugs should have been 'critical'. I'm going to propose
a new system for releasing things into stable that will solve these
problems for the next major release.
> What's the focus here? Is debian trying to be a
> solid system or a cutting edge system.
Both. But the 'stable' release should focus on being solid - and the
'unstable' working version should focus on being cutting edge. That
way, somebody that is sticking with 'stable' will be assured of having
a solid system that is only 6 months behind the 'bleeding edge'.
Think of most of the Debian developers as people who are willing to
live with the 'bleeding edge' and try to work out the bugs. The 'unstable'
distribution can be quite unstable at times (I lost the data on my
hard drive to one bug).
Stability and being on the 'cutting edge' are mutually exclusive -- until
a lot of work is done to fix bugs. We have 200 developers to do the work,
which might sound like a lot, but it isn't. Nobody's doing this
full-time as far as I know (although it may seem like it).
> Paragraph 4 of the Debain
> Social Contract ("Our Priorities are Our Users and Free Software")
> states (whole P not quoted),
>
> "To support these goals, we will provide an integrated system of
> high-quality, 100% free software, with no legal restrictions that
> would prevent these kinds of use.
>
> I find that while the quality of distributions is generally "high"
> the emphasis from the developers is more on "let's get that hot new
> release going" rather than "let's get all the bugs out of this latest
> distribution and make it completely solid".
That rings true. Basically 'stable' is a snapshot of 'unstable' at a
point in time where the developers have all taken a hiatus from adding
new features, and concentrated on fixing bugs. Overworked (or absent)
developers might not be able to participate fully in the release
effort, and we only have cursory checks (Dale Sheetz and Brian White
+ their helpers), so bugs slip through.
I don't think we have any developers actively working to improve
'unstable'. That's a good thing - who'd do the testing?
I'll admit that our testing/releasing procedures could (and will) be
improved. Debian 1.3 is really only the 3rd major release we've done
(my 2nd). We're still learning. Being the distributed bunch of
volunteers that we are, we don't have the luxury of doing testing in
the conventional top-down, heirarchial approach (which takes considerable
resources). We have to be creative. I think it will take several more
iterations before we get it 100% right, and no more bugs slip through.
> If this isn't what debian's about then I'm sorry for making all this
> trouble. I'll shut up and go get Caldera. It's just a shame that the
> distribution which I think has the most going for it technically and
> produces a distribution so easy to maintain can't achieve this level
> of quality.
Once we get the proper procedures in place, I have no doubt that the
quality of our distribution will be the best. We're too dedicated to
the effort for anything less. :-)
> Alright, that's just about enough out of me.
Quite alright. I think we developers appreciate the feedback and
need it (even if it isn't always good news). Thanks.
Cheers,
- Jim
Attachment:
pgpHn4yIVknSI.pgp
Description: PGP signature