[repo-coord] kernel module versioning (was: Blackmailing ...)

[Dag Wieers]

On Thu, 6 May 2004, Axel Thimm wrote:

> On Thu, May 06, 2004 at 01:25:23PM +0200, bent at biosci.ki.se wrote:
> 
> > To clueless me it doesn't sound terribly broken....
> 
> It isn't. You are confused, because both start with "kernel-module-",
> but the one Dag promotes has the kernel data in rpm's release tag, not
> the name tag. This severely breaks any upgrade paths as discussed in
> these old threads above.

It doesn't break the upgrade path, and you seem to like 'overstating' 
things by using wordings like 'breaking severly' etc...

It isn't a very good solution because it doesn't tend to automatically 
update kernel modules _without_ changes to apt or yum, or without a 
helper. You have to do it manually, which isn't that hard but needs 
attention when upgrading the kernel. Not the ideal situation. Not want 
end-users expect anyway.

I've always stressed this, I'm not denying that. Axel's strong wordings 
make it as if we're denying it and don't want a working system or do this 
on purpose.

Like Panu says, it's not very sane because name-tags were never intended 
to have versions, but out of practical considerations and because there is 
no alternative I fear we have to do this.

Given the fact that both Matthias and I work together closely and are 
building from the same base, I depend on a consensus to move over to a new 
versioning scheme. I can't just make that decision although I'm already 
convinced there's no alternative. And no new alternatives on the horizon.


> The kmdl vs kernel-module, or module name before or after these is
> a purely non-technical issue.

Sure, the naming-issue is not tight to the versioning. And the versioning 
is not tight to obsoleting uncommon names. But everytime I talk about 
moving to the kernel-module-standard, you brought the versioning up 
because you know we've got issues with that. So you just bounce the ball 
back to us. Which is part of the impasse.

And I do tight the kernel-module naming together with obsoleting uncommon 
names because they are tight. If there's no naming standard, things may 
still break (you couldn't be installing a module I have and a module Axel 
has).

The reason why it is important to have this all straigthened out together 
is because with 4900 kernel-module packages I'm not going to recompile 
them now to fix the obsoletes and then do it all over again because the 
rest wasn't properly standardized.

Unfortunately Axel calls this blackmailing, I call that common sense.

I just wished Axel wouldn't be so harsh in these discussions and keep it 
to the essence. Despite all that I hope we might have a final solution in 
the making.

--   dag wieers,  dag at wieers.com,  http://dag.wieers.com/   --
[Any errors in spelling, tact or fact are transmission errors]