Re: debian experimental == ubuntu hoary?
Marcelo E. Magallon wrote:
Hi,
I agree with Anand's POV.
I think I have stated a couple of times that I dislike the kind of
abuse of 'experimental' that the gnome packages do, mostly because it's
working around problems with the 'testing' idea (a package can
percolate from unstable to testing when it actually shouldn't because
of unexpressed dependencies -- i.e. an actual bug in the package that
can't be catch by the testing scripts).
I agree this is a bug - but I don't think staging in experimental hides
the problem. If you then upload the whole GNOME release en masse to
unstable, they can still propogate piecemeal to testing as
Recently I realized that this
is also working around the NEW queue problem: by staging packages _in_
the archive (e.g. in experimental) you basically put a package in the
queue for ftp-master's attention and by prodding some ftp-master (say,
on IRC -- and this is a discussion by itself) you can skip ahead on
that queue.
I think implying that GNOME packages somehow obtain preferential
treatment from FTP masters by prodding them on IRC is pretty far out.
Most of the GNOME packages are already there in the overrides file, and
doing your time in queue/new waiting for acceptance into experimental is
still doing your time. It just means that when you upload to unstable,
you can do it all at once, which is better.
Its not cheating the system - you can do it too. I uploaded new split
gaim/gaim-dev/gaim-data packages to experimental and had some people
have a go with them (particularly people who wanted to package gaim
plugins), fixed a few bugs, and when a new upstream release came out,
they went into unstable straight away in a far better state than they
would've otherwise been, with no delay.
What I'm saying is that I don't see an actual technical reason not to
use a staging area *outside* the archive (say, alioth) since -- at
least the last time I read about this -- experimental isn't
autobuilded. One could argue about mirrors, but my *feeling* is that
the kind of people who are willing to use packages from experimental
don't mind much about not having certain package mirrored. I'll be
delighted if someone comes up with transfer statistics that prove me
wrong.
Experimental is in the process of having autobuilders arranged - it's
also common for developers of other popular desktop architectures to
upload packages there. I don't see the problem
I actually like your argument for piecemeal upgrades. If the package
interrelationships do not express a dependency, piecemeal upgrades
should work. If they don't, that's an RC bug by our own standards.
I agree. But I also agree that in general, we should heed upstream's
advice of staging GNOME releases together, because they tend to work
well that way, and are often richly interdependent anyway (new
components depend on new libraries etc).
On top of all that, using Ubuntu as a staging area for Debian isn't one
iota better. Debian users shouldn't have to go cherry picking to
Ubuntu's archive. Solving Debian's problems *outside* Debian does not
help Debian. It's rather easy to figure that one out, isn't it?
That's just bull - since *before* Ubuntu turned up, we have been staging
only even-numbered stable GNOME releases in experimental. Just because
some of the people who do this are also maintaining
unstable/preview/beta releases in Ubuntu doesn't change this, it just
means that we'll probably be able to get the new release into
experimental faster. So we're not missing anything we had before, and
we're getting what we want (or at least, what the release managers want,
and I think they're right) faster. Where's the problem?
Marcelo
Regards,
Rob
Reply to: