Two theses regarding packages

[Adrian Bunk <bunk@fs.tum.de>]

Hi,

below are two theses (with explanations) that reflect my opinions and I'd
like to get some feedback whether other developers agree with them or why
they disagree:


1. The maintainer is responible for the quality of his packages.
He should try to fix bugs within a reasonable amount of time.

2. When we consider something important the policy should force
maintainers to change their packages.



Explanations and remarks for thesis 1:
- "try to fix bugs":
  I expect a maintainer to be able to fix a "mising build dependency on
  flex" bug but it might be that e.g. the gcc maintainers can't fix an
  "internal compiler error" on ARM.
- "reasonable amount of time":
  Most of us work for Debian in our spare time. But I think we can expect
  any maintainer to try to fix at least all RC bugs within 1-3 months.
- I consider sending a RC bug not "asking for the removal of a package"
  but "asking the maintainer to fix it".
- When a maintainer doesn't try to fix his packages we should reconsider
  whether he's really an active member of Debian.
- The point above is only a last refuge (and it's already accepted
  practice that tbm marks packages of extremely inactive maintainers as
  orphaned).


Explanations and remarks for thesis 2:
- This is an opposite to "the policy should only document current
  practice". Let me make a current example (that was one of the reasons
  for this mail):
  I think we should (soon after woody is stable) make it a "must" in the
  policy that packages that need packages to build that aren't installed
  on a system with only the essential and build-essential packages installed
  need to declare build dependencies. A "let's wait until only 200
  packages are missing build dependencies" approach misses e.g. one
  important thing: Some maintainers start thinking about removing the
  build dependencies from their packages - why shall I have build
  dependencies in my packages when they only cause RC bugs because of
  missing build dependencies? Noone can stop me when I remove the build
  dependencies from my packages (and I think the release manager will
  lower the severity of all RC bugs because of missing/removed build
  dependencies because there's only a "should" in the policy...).
- It's clear that today (hopefully) not too far away from the next freeze
  it's not the right time to change the policy and to send RC bugs for
  e.g. the lack of build dependencies although they are needed, but I
  consider it a good idea to change this soon after woody is stable (and I
  have no problems with having many RC bugs because of this - according to
  thesis 1 I expect the maintainers to fix them soon enough before the
  next freeze).


Yust my 0.02
Adrian

-- 
A "No" uttered from deepest conviction is better and greater than a
"Yes" merely uttered to please, or what is worse, to avoid trouble.
                -- Mahatma Ghandi