Re: Redundant fields in debian/upstream/metadata and possible lintian check
Hi Andreas,
thans for raising the issue. In brief, there have been unsuccessful
experiments, but I think that we should let people experiment.
Le Wed, May 23, 2018 at 03:39:44PM +0200, Andreas Tille a écrit :
>
> at least in scientific packages the file debian/upstream/metadata is
> frequently used since it is the established way to specify citations
> belonging to some software.. The definition of the fields is given in
> Wiki[1].
Side note: while in practice, multiple people are subscribed to
notifications of changes of this page, which means that non-consensual
additions would be quickly spotted, I think that it would be good to
clearly mark which fields are broadly used and accepted, and which are
more experimental or anectodical.
> These data are gathered in UDD[2]. When I inspected the log of the UDD
> importer I noticed that there are a lot of redundant fields like
> "Homepage" or "Watch" where we agreed that these fields should not be
> duplicated in upstream/metadata.
Indeed, I was hoping that they could in the long term supersede the
Homepage field of debian/control and the debian/watch file, but it is
not going to happen anytime soon. I think that it would have been nice
to be able to propagate updates of this information to the Debian
infrastructure without doing a package upload, but... I guess that
somebody else will eventually find a better way to do this.
> There are also typos and freely invented Fields which are not
> specified on Wiki[1] (like Distributor', 'CRAN', 'Wiki'). I think it
> makes sense to have some lintian check for this undefined fields. I
> think I'll file a wishlist bug about this soon.
I think that it is important to let people experiemnt and introduce new
fields. Nevertheless, typos and fields with too similar names should be
prevented. Maybe a Lintian check could send a warning for any field
that has an edit distance of 1 compared with all the "broadly used
and accepted" fields ?
> However, before I do I'd like to discuss the fields Name and Contact.
> DEP8 defines[3] the fields Upstream-Name and Upstream-Contact which are
> the same values in a file that has a high probability to be properly
> maintained.
(You mean DEP 5). Since the upstream contact address is volatile, I
think that ideally it would be better placed in the upstream/metadata
file. But that may require a change of Policy and other potentially
very painful discussions...
> In the case of r-* packages from CRAN or Bioconductor it can be even
> automatically updated (via dh-update-R ... its actually not really
> done but I think this could be implemented easily - dh-make-R at least
> generates the fields at the time of initial package creation).
>
> I wonder whether we should maintaining duplicated information and thus
> would like to hear your opinion about orphaning these fields in
> debian/upstream/metadata.
Homepage and Watch were already removed from the wiki page some time
ago, but could come back into an "actively deprecated" list of fields.
I think that any duplication done by hand for a long time is going to
create noise and cost time. But duplications made by automation tools
such as dh-update-R are potentially useful and sould be more considered
as "synchronisations" which propagate information from external source
into the Debian infrastructure.
Have a nice Sunday,
Charles
--
Charles Plessy
Debian Med packaging team,
http://www.debian.org/devel/debian-med
Tsurumi, Kanagawa, Japan
Reply to: