Re: Bug#684396: ITP: openrc -- alternative boot mechanism that manages the services, startup and shutdown of a host

[Russ Allbery <rra@debian.org>]

Philipp Kern <pkern@debian.org> writes:

> Of course, if GNOME is unused one could just remove it completely from
> those ports, but I doubt that your approach of "it's just a minute of
> work to RM it" is welcomed. (Well, the maintainers would probably like
> it, as long as there won't be bugs claiming that you have to support
> that port because it's a Debian port in testing.)

It probably works okay for the kind of packages that Charles is primarily
talking about, though (scientific computing packages written by scientists
rather than software developers that have probably never been run upstream
on a non-Intel processor and are usually leaf packages or, if not leaf,
only depended on by other packages with a similar profile).

We don't have a particularly good way of handling this situation right now
other than one-off work on each package that may need to be treated
unusually.  It's a bit difficult for the maintainer to determine the
implications for the dependency graph, and there isn't any good way to
exclude all packages in a particular class from a particular architecture.
We have some architectures where I really doubt that anyone is using them
for anything other than a server (s390, for instance), and (modulo cases
where it makes sense to run such software as part of a remote session on a
shared-user system) we don't have a good way of easily flagging local
desktop software that probably doesn't make a lot of sense in that
context.

That said, we do get value from porting that software to all
architectures, even if it's unlikely anyone would want to run it there.
Several of those architectures have unusual features that have later
turned up in mainstream architectures, and the earlier porting efforts
have given Debian a huge leg up.  For example, our 64-bit porting was
mostly already done by the time that amd64 became a popular architecture,
thanks to other architectures that people would have considered obscure.

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>