Re: stable / backports (Re: When Debian 4.1 will arrive... will anyone care?)
On Fri, Apr 20, 2007 at 08:11:20AM +0200, Bart Martens wrote:
> On Fri, 2007-04-20 at 15:17 +1000, Craig Sanders wrote:
> > contrary to popular belief and self-delusion, 'stable+backports' is NO
> > LONGER STABLE.
> That is of course true.
> > the only 'advantage' to using 'stable+backports' over 'stable+some
> > packages from unstable or testing' is that you don't have that nasty
> > label 'unstable'.(...)
> > IMO, if you need a 'stable' system with some newer packages, you're
> > better off learning how apt's pinning stuff works than bothering with
> > backports. it's not hard.
> Backports are recompiled packages from testing, so they will run without
> new libraries on a stable Debian distribution. It is not always
> possible to install a package from testing without pulling in lots more
> packages from testing.
> But, as you know, stable+testing is no longer stable either. :)
yes, i know. that's exactly my point. that there's no difference
stability-status-wise between using stable+backports and stable+testing
or stable+unstable. once you've installed anything from any of them,
you're no longer running 'stable'.
> > to get that crucially important 'benefit', you're using packages
> > from a repository with unsigned packages by unknown maintainers.
> Last time I checked, only DD's can upload to backports. But it is
> correct that anyone can create a package for backports, and ask a
> DD-sponsor to upload to backports, without consulting the maintainer
> of the package in testing. So this seems similar to an NMU. That
> introduces a slight risk that some details might be overlooked.
it was my understanding that while some backports packages are
maintained by DDs, sometimes even the same DD as for the debian archive,
not all of them are. i.e. packages in backports could be made and/or
uploaded by pretty nearly anyone.
> Anyhow, I think that the discussion is about getting newer upstream
> releases into stable sooner. The backports project is an interesting
> approach, because it makes newer upstream releases available to stable
> users, without putting these packages in the Debian stable repository.
> Somehow I think that the stable-ness of Debian stable is one of the
> strong assets of the Debian project, so policy about how stable a
yes, it is.
i just don't see why people like to fool themselves that they're still
running 'stable' when they install stuff from backports. they're not.
i think people are needlessly scared off by the name 'unstable' and, to a
lesser extent, 'testing'. the implication is that the distribution itself is
unstable or flaky....when, in reality, what 'unstable' means is that it is
constantly changing. this may occasionally cause problems, but generally
doesn't....and most of them can be avoided by running 'testing' instead of
i don't see the need for backports. stable is good for those who
need/want a stable environment that doesn't change much except for
security updates. testing and/or unstable for those who want bleeding
edge. stable + apt pinning + either testing or unstable for those who
want something in between.
i run a mix of stable and unstable servers. predominantly unstable.
have done so for over a decade. truth is, i have more problems with my
stable servers because i can't provide the same stuff on them (e.g. new
versions of packages like php, or packages that aren't even in stable)
that i can on unstable - if it can be done at all, it's a lot more work
and hassle than it is on the 'unstable' machines.
> package in stable should be, should not be changed too sudden. A
> gentle transition to any direction should be OK, as the project
> should be allowed to evolve. Let's not confuse this with getting the
> packages in unstable/testing updated to the newest upstream releases
> some time before the next freeze preparing the next stable release.
i agree. i think stable should be kept stable. unstable and testing is
where the new/updated stuff belongs, not in stable.
backports is just an unofficial variant of 'unstable' by another
name...and anyone who's scared off by the name 'unstable' should be
equally scared off of 'backports' (or more so).
craig sanders <email@example.com>
BOFH excuse #251:
Processes running slowly due to weak power supply