Re: A concrete proposal for rolling implementation
On Wed, 04 May 2011, sean finney wrote:
> It's an excellent idea. Some of the initial feedback that I've gotten
> about DEP-10 (in particular some brainstorming on IRC with Carsten Hey)
> is pointing at ideas along these lines, and I hope to flush them out
> in a bit more detail RSN. But I think it's particularly exciting that
> these two ideas (rolling, and dealing with freezes) might not conflict
> with each other, and perhaps complement one another.
> One issue that would need to be addressed with experimental is that
> opening a migration route anywhere out of experimental might come as
> an unpleasant surprise to some, since there's a standing expectation
> that it's a pseudo-suite where we can put stuff that we don't necessarily
> want to try out in unstable. Not an insurmountable problem (either we
> change that or introduce yet another psuedo-suite for this purpose),
> but worth note anyway.
Yeah, experimental is not really the good place. We really want in
rolling only packages where we have the assurance that they will land
in unstable the day after the release (so automatically and not with
a manual source upload).
So I'd favor some sort of unstable overlay that is not experimental.
It could be called "unstable-next" and could be auto-generated from
uploads targetting unstable that introduce a new upstream version.
That way by default unstable doesn't move forward with new upstream
version and can always be used to upload bugfixes targetting testing.
Auto-building in unstable-next would be like experimental, i.e. it
would be built in unstable so that it's still possible to pick
packages there in the rare case where a new upstream version is
desired late in the release cycle.
I like where this is going! :)
Raphaël Hertzog ◈ Debian Developer
Follow my Debian News ▶ http://RaphaelHertzog.com (English)
▶ http://RaphaelHertzog.fr (Français)