[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: The stable/testing/unstable branches not a solution ?


I think that this conception in itself is just a workaround of the
fact that the release cycles are too long. If they were shorter, then we would
not have a very old stable. 

The approach to QA work should definitely be reconsidered, a new
approach might contain some changes to the branching, but I
think it should be mainly concerned with the following areas:
-release goals
	There should be small release goals. One goal for
	one release ideally. As a barter, we should set the
	release goals for more releases in advance, of course
	the goals for releases and their sequence could be
	debated any time for anything not the next release.
	It also means that a release often means just
	letting already implemented features into stable.
	Your proposal comes into play here. See below.
-hard release measures
	Any packages with an RC bug should be dropped from 
	the list of candidates for the next release until
	the bug is fixed. Also the approach of NMUs might
	have to be changed.

Preparing to some releases (even to future releases) some maintainers
will create packages and package clusters they do not want to
include in the release for some reasons, but they want to see them
in sid. A solution to this could be a For-release: control
field containing the code name of the future release where one wants
it to be included. This means that the code name and the release goals
of the release should be undetachable (this not means that changing the
details of the release goal is prohibited), but the sequence of them is
still open to discussion.
The automatisms could decide based on this header on which package goes
into the release candidate.

A levelezőm azt hiszi, hogy Lucas Nussbaum a következőeket írta:
> However, I remember that Raphael Herzog said something very interesting in his platform for the 2002 DPL election[1]. He introduced the concept of a "working" branch, where packages would go when the maintainer ask for after they would have been tested enough.
> [1] http://www.debian.org/vote/2002/platforms/raphael#release1
> I personally don't think Debian should ever release real new versions. Snapshots of a working branch with not so outdated software would be a much better solution. Most of the potential Debian users have broadband internet access now, and don't care about a 7-CD set. Of course, my opinion doesn't have much value, since I'm not even a DD, but I would be very interested in your opinions.
> Thank you,
> Lucas
> -- 
> To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

GNU GPL: csak tiszta forrásból

Reply to: