Re: Added README.source to our group policy.
Am Donnerstag, den 13.11.2008, 19:47 +0900 schrieb Charles Plessy:
> Le Thu, Nov 13, 2008 at 10:42:37AM +0100, Daniel Leidert a écrit :
> >
> > IMHO this is a misuse of this file. The contents of README.source should
> > point to unusual/uncommon source handling, including source parts, which
> > have been removed, repackaged or put together, patch management systems
> > and maybe information related to multiple/special builds. But your
> > template does not fit any of these categories. Instead it doubles the
> > information from the already existing Vcs* control fields and
> > debian/copyright. So I think, that the last paragraph of section 4.14 of
> > the policy doesn't apply here (IMHO *buildpackage tools are not the
> > target of this sentence, as they are not necessary to modify the
> > source). As long as this package is maintained by a group of people,
> > they can agree to some policy internally. But such a group policy nor
> > the VCS usage do not apply to NMUs nor new maintainers (if this package
> > gets orphaned). So I think, putting this information into README.source
> > is wrong.
>
> Hi Daniel,
>
> In theory, nobody should NMU Debian Med's packages, as we are very active and
> usually repsond to emails within a day.
How many 0-day NMUs did you observe (except in release times)?
> In practice, self-appointed "NMUers" do not care and rush on low-hanging
> fruits, introducing discrepancies between the Debian archive and our
> repository and elevating our work load for no global benefit.
It is well known, that NMUs should be uploaded in an x-days queue with
x!=0. However, this is not argument against my point of view. If you
feel, that 0-day NMUs increased too much, then this is something, the
project has to discuss and solve. This will help you much more than
putting a paragraph in an unrelated file.
> I therefore think that it is necessary to document somewhere:
>
> - That a package is group-maintained and that it follows some packaging rules.
Vcs*, debian/copyright, debian/control - there *are* places to document
this
> - That all NMUers have write access to our repository and we expect them to
> commit their changes if they are so disappointed by us that they think that
> the only way to solve a problem is to ignore us and NMU the package.
That's a bad claim. You have no right to force people doing QA work to
commit to your SVN. Please suggest this to the QA team. I don't think,
they will accept your request. It's not their job to care about your
group policy nor your VCS. Further usually you have several days to
prepare your own update, so the NMU will never reach the archive.
> So yes, README.source is supposed to be useless regardless of its content for
> packages that are actively maintained,
Also "active" maintainers sometimes have vacation, are ill or have other
reasons to be unavailable to fix an RC bug. Or they drop interest in a
package. And exactly for these reasons, README.source should document
the source handling. And this doesn't need information about groups and
their Vcs.
> How about the following modification:
[snip]
I don't agree to your usage of README.source. I still think, that it is
better to create 2 new fields: Vcs-Group and/or Vcs-Policy if you really
think, it is necessary to document this somewhere. But README.source
IMHO is the wrong place. Putting such information into it, will make it
harder to get the really necessary information about the source from
this file.
[..]
> It only invites to add the template in the cases where README.source is
> recommended by the Policy,
But you don't use the file as intended in the policy. That's what I
criticize.
[..]
> If other members disagree that a commit would be welcome (bear in mine
> super-busy NMUers do not event bother sending debdiffs to the NMU'ed bug when
> it is too trivial, which is of course the case for low-hanging fruits), I can
> remove the paragraph. But personnally I will keep it in the packages I consider
> myself as main responsible.
And because of this you are listed in Maintainer or Uploaders. No reason
to put this information into every other file as well. The mentioned
fields are intended to list the responsible person.
PS: I would not be that critical with NMUs for trivial changes. You are
not forced to have a copy in the tags-directory of your SVN. In
question, I would simply copy the changelog entry from the NMU into the
SVN and make the change myself (especially if I don't agree how the
NMUer fixed the problem). No need to tag the NMU package version.
Regards, Daniel
Reply to: