[Pkg-pciutils-discuss] Re: Bug#328318: pciutils: why not centralize pci.ids?

[Matthew Wilcox]

On Thu, Jun 15, 2006 at 10:05:07AM +0200, Cristian Ionescu-Idbohrn wrote:
> On Wed, 14 Jun 2006, Matt Taggart wrote:
> > Have you tried /usr/bin/update-pciids from the pciutils packages? :)
> 
> No, I did't :( Another one I missed. Thanks for pointing that out.
> 
> > The drivers in the Linux kernel all have their own lists of PCI ids
> > that they support. I don't know what kudzu/etc do, but I didn't
> > think their ability to support hardware was based on how up to date
> > their pci.ids was.
> 
> I suppose (didn't look into the details) discover, hotplug, udev, etc.
> have their own way too.

They don't rely on pci.ids for naming, no.

> > Anyway we'll try to keep it as up to date as possible. After a
> > stable release happens I suppose we could update via a stable update
> > if needed, but we'd need to have a good reason.
> 
> Have a look at the 'hinfo' package. When installing, it offers to
> setup a cron-job.

I don't think this is a good idea.  It's going to put a lot of load on
the sf.net webservers to have a few million Debian machines all trying
to update at the same time.  It also messes up debsums (and no, I looked
at faking out debsums ... that's not a good idea either).

If we're going to update pci.ids, I think we should use volatile for
doing so.  Maybe we should split out pci.ids from pciutils for that,
since it would then be binary-all.  But I'm not too keen to proliferate
packages unnecessarily.