Re: Evince backport proposal

[Sylvain Beucler <beuc@beuc.net>]

On Wed, Jun 14, 2006 at 10:34:35AM +0200, Daniel Baumann wrote:
> Sylvain Beucler wrote:
> > I made a backport of evince 0.4
> 
> can you please backport the testing version, we don't want unstable
> backports on backports.org (except for security updates of already
> accepded backports). Alternatively, you could wait until the -2 enters
> testing.

What is the rationale behind this policy? It seems strange:

- Debian packagers are encouraged to and do use official tarballs from
  upstream rather than VCS development version (the use of "unstable"
  to describe sid refers more to the dependencies than the software
  per se).
  sid offers the latest releases, and that's what we aim at.

- "testing"'s goal is to automate some dependency consistency checks,
  but that's irrelevant to bpo, because bpo is primarily about
  altering those dependencies.
  Dependency checks are manually performed by each backporter.


Sticking to testing already leads to contradicting effects:
backports.org offers a buggy version of OpenOffice2 (2.0.1) from
testing while there is a bug-fixed version (2.0.2) in unstable for
months. I couldn't install it for my clients because 2.0.1 freezes
with NFS, and I had to fall back to alien-ing the official 2.0.2 RPMs
(where NFS locking is fixed)...

'unstable' is more stable than 'testing' in this case.


As a result I'd intuitively tend to use sid in most cases.


In the particular case of evince, the difference (dependency fix) is
so meaningless that it indeed doesn't matter, but I'd like to better
understand this first.


Cheers,

-- 
Sylvain