Re: The nature of unstable (was: Danger Will Robinson! Danger!)

[Edmund GRIMLEY EVANS <edmundo@rano.org>]

Paul M Sargent <pauls@3dlabs.com>:

> OK, Here's a question then. If Woody is unstable, which kernel is it
> running?

> Woody should be running 2.3 or pre-2.4. That should have been among the first
> things to change.

I don't think so. People who are interested in debugging the kernel
can install 2.3 themselves; people who are only interested in
debugging Mozilla (say) don't want to have new kernel releases
trashing their discs every few days.

> On a side note. I'm really not sure that this 'release' stuff works on
> debian. Coordinating the development cycles of an infinite number of
> packages is impossible. What I would like to see is an unstable tree where
> all development is done. As packages reach maturity they 'graduate' to the
> stable tree. A snapshot of stable tree at any time works. The unstable tree
> just becomes a place for developers to share packages.

I made roughly this point in a message that seems not to have reached
the list. Where there are no complex dependencies, there is no need
for packages to be released (i.e. declared "stable") simultaneously.

I suggested creating a "stable-test" branch containing packages that
are built to run on "stable" without breaking anything in stable or
stable-test. When a package maintainer is confident that the version
in stable-test is in every way superior to the version in stable he
asks the release manager to move that package from stable-test to
stable. (I'm not sure how the version numbering would work, but no
doubt a numbering scheme could be invented.)

Edmund