Re: [gentoo-dev] Re: rfc: revisiting our stabilization policy

[Rich Freeman]

On Sun, Jan 26, 2014 at 1:56 PM, Peter Stuge <peter@×××××.se> wrote:
>
> I don't think that's "completely optional" though, it sounds like a
> one-way function. If have ever stabilized a package once then must
> ensure a stable package forever.
>
> I think arbitrarily removing stable versions should also be an option,
> and I think package managers would be able to deal with that without
> much extra effort?

Well, I think we certainly should be able to de-stabilize packages.
I've seen this happen in the past, especially when the need to not be
stable is inherent in the package itself (such as game clients that
need to be synchronized with servers - only one version will ever work
at any time buggy or not).

Ideally this should really be the result of communication between the
maintainer and arch team.  What we definitely don't want is a package
that gets stabilized, and then six months later the whole package is
back at ~arch, and then six months later there is a stable version
again, and so on.  That just isn't, well, stable.

However, if an arch team is feeling overwhelmed I'd strongly encourage
them to put out a bulletin telling maintainers to stop STABLEREQs for
non-system packages, or whatever other guidance they want to issue.
It has been pointed out that on these archs system packages often
don't work, so having those be stable at least lets them target
versions they want to fix up and lets users get a bootable system
without too much fuss.  Falling back to a defensible position and all
that...

But, nobody needs anybody's permission to do any of this.  Ideally the
arch team should take the leadership to establish a policy on their
arch which is maintainable.  If they don't do that, well, then
maintainers complain and we get threads like this one.  The arch team
has the greatest interest in having the arch work - I'd strongly
support them in creating any policy for their arch that they can
follow-through on.

Rich