Re: debian experimental == ubuntu hoary?
On Sat, 05 Feb 2005 16:28:37 +0000, Robert McQueen wrote:
> Anand Kumria wrote:
>> That's really a maintainer decision to make.
>
> Almost all of the core GNOME components are maintained by the GNOME
> maintainer team, so they will tend to act in a unified manner.
Oh, I actually thought the team was equal to Sebastian completely. Sorry,
I wasn't aware there was more than a single person involved.
> What I have
> described is their current policy, which is generally supported by the
> release managers.
As I mentioned, the release managers required the gnome piecemeal
transition to function correctly -- at the moment you have upstream
(GNOME) saying that they don't support mixing and matching (w.r.t Jeff's
recent email).
And you also have Debian saying piecemeal upgrades (via a via the release
managers) are a must. The Gnome team handle that impendance mismatch by
staging the release via experimental.
I believe it'd be best to not keep that mismatch hidden away in
experimental but instead make it directly visible in unstable.
[ discussion of unstable, etc. ]
Basically I completely disagree with you. In fact the fact gtk+2.6 is in
unstable means that some of my packages have already broken. So, I don't
even think that the Gnome maintainer agrees with you in principle.
>> That kind of problem is something that is well hidden from other
>> developers and really destroys the utility of having 'unstable' in the
>> first place.
>
> People who maintain GTK+/GNOME based packages are exactly the people who
> should pay attention to the fact that GNOME releases are put into
> experimental first. That's why this mailing list exists, and it's
> announced and co-ordinated here.
Please. Next you'll be saying I have to listen in to debian-kernel to know
when an interface has changed. It should just go into unstable, so I can
observe the breakage directly.
>> To me a GNOME release is a "we believe these packages work well
>> together". I don't think it should necessitate them being packaged /
>> uploaded all at once.
>
> We should upload it when upstream calls it stable, and GNOME does that
> with the components of their desktop all at once. If upstream don't
> believe they work well together... surely that's a good hint not to put
> them in the critical path for our release?
You earlier assessment that preventing transitions from unstable into
testing are pretty much correct -- having to file an 'RC' bug is ugly.
However this idea that we should only upload when upstream says doesn't
fly. If every maintainer followed the same line of thinking there would be
plenty of software never packaged -- quite a lot of 'upstreams' never
quite consider there software finished.
Our task is to decide when it is good enough for a wide audience (via
unstable) irrespective of the version number / label that they might put
on a release.
My criteria is (generally): has the maintainer released a version? If I
can't find a show-stopper, upload it. I don't believe a "blessed" set of
packages should prevent the uploading of individual components.
Cheers,
Anand
Reply to: