Re: post-release package update policy
firstname.lastname@example.org (Ian Jackson) wrote on 23.10.95 in <m0t7M4E-0002b8C@chiark.chu.cam.ac.uk>:
> This is missing the point. I think that trying to impose a version
> number on the whole system is a *very* bad idea. It will artificially
> tie us to this absolutely ghastly idea of having frozen snapshots.
Well, the problem is that Debian is already doing this, announcing 0.93R6
and talking aboutwhat 1.0 will be.
I agree, however, that I have some trouble seeing the point of such
releases as it is today.
Maybe we should take some advice from the way the kernel is handled. It's
already partly there, with the "experimental" hierarchy.
How about the following scheme?
debian-known-good/ Packages that have no serious bugs left.
They shouldn't be moved here before they've aged,
say, a month or two.
debian-newest/ where most packages go after they arrive in
Incoming. They get removed when proven bad, or when
moved either to or out of known-good - that's a
matter of taste.
debian-experimental/ just what the name says.
debian-historic/ Packages moved out of known-good, for as long as
there is space enough.
That's only a sketch, of course, but it should convey the general idea.
Further, at the time of switching to ELF (which I'd define as the time
when we can have a meaningful ELF known-good directory), we could make a
debian-known-good-aout/ and a debian-known-good-elf/.
> When you report a problem you just report the version numbers of the
> relevant packages on your systems - including the base disks, if
> relevant. I think that the version number of the base disks used to
I think an easy way to include a list of all installed packages with
versions in a bug report should be in the system, and *documented* clearly
enough so it will be used.
Listing only the relevant packages is fine, but sometimes you don't know
which are relevant.
Suppose elvis installs a broken command over one of emacs. You have long
since forgotten about this (if you ever noticed) and report the bug as an
emacs bug, without including elvis version info ...