Re: Bug#645656: network-manager in Gnome
Stefano Zacchiroli writes ("Re: Bug#645656: network-manager in Gnome"):
> On Tue, Nov 01, 2011 at 01:42:47PM +0000, Ian Jackson wrote:
> > Yes, that much is obvious. So the answer is that there is no harm in
> > actually removing network-manager.
> There is harm in diverging from upstream. We're a software
> distribution, by default we integrate existing software and we try to do
> so staying as close to upstream as possible. We do diverge from upstream
> when we've good reasons to, but we don't do that "just because".
We should do it when we judge that the benefits are worth the costs.
In this particular case the costs seem to be minimal. There isn't
even a direct patch-carrying cost, since the dependency is expressed
in our own control files.
It's not clear to me that the Debian maintainers have made this
judgement on the basis that I would be expecting, perhaps because they
don't see that there is a problem.
> Given you have a way to install a GNOME environment without n-m, I fail
> to see your point. At best, it seems to me you should be arguing with
> the GNOME maintainers to have another meta package ("gnome-minimal"?)
> and respect their choice if they don't want to.
AIUI previously the "gnome-core" package served this purpose. I don't
know what proportion of users who have "gnome-core" but not "gnome"
installed have done so to avoid network-manager (and perhaps other
things which have been declared upstream to be part of the core). If
a substantial number of them did then adding n-m to the dependency set
is quite unfriendly.
#rank name inst vote old recent no-files ...
908 network-manager 43263 34749 5170 3337 7
989 gnome-core 40526 1 0 0 40525
1395 gnome 29256 0 0 0 29256
1561 kdelibs-bin 22638 12002 5792 4841 3
2784 xfce4-panel 8071 3258 3248 1561 4
This doesn't really answer the question unfortunately.
I'm not adamantly opposed to the idea that there should be some other
metapackage, rather than gnome-core, to serve the need of those
users. But I think we need to have a proper conversation about it
rather than just dismissing the proplem out of hand.
> FWIW, I don't think that escalating this to -devel in search of
> pitchfork equipped network-manager haters is a good strategy either.
I think escalating a bug to debian-devel is a perfectly reasonable
approach. The fact that many other people are likely to be interested
is /more/ of a reason to use this forum, not less of one. At the
moment I get the impression the maintainers don't feel that there is
actually a problem, and I think canvassing opinion here is one way to
demonstrate that there is. Or alternatively confirm that there isn't.