[ATrpms-devel] Re: Changing foo-kmdl- to kernel-module-foo-? (was: Blackmailing ...)

[Axel Thimm]

On Tue, May 11, 2004 at 05:55:59PM -0500, Gary Lerhaupt wrote:
> On Tue, 2004-05-11 at 03:06, Jarod Wilson wrote:
> > Sorry, been stacked up w/work and personal stuff for a little while 
> > here, but I'll toss in my worthless two pennies before finally getting 
> > to sleep...
> > 
> > Personally, I like the foo-kmdl convention much better. I like the 
> > kernel module listed along with the other components (a la Synaptic). I 
> > do see the benefit of all kernel modules being listed together, but 
> > kernel-module-foo is a ton of extra characters tacked onto an already 
> > likely very long name. However, for the sake of consistency and 
> > compatibility between all repos, it is probably best to give in and 
> > make everything uniform (pick your battles, this one probably isn't 
> > worth fighting). It sucks, but we'll get over it (whoever "we" is!).
> > 
> > Time for bed, throat is hoarse from shouting. ;-)
> 
> Sorry for joining the discussion late, but I've been discussing this
> this with Axel off the list.  Collectively, have you considered using
> DKMS to handle all this naming/versioning mess?  Rather than packaging
> and repackaging for every kernel version, with DKMS you just release
> once per module version and let DKMS recompile for whatever kernel the
> end user is using.  It also allows you to include pre-built binaries for
> commonly used kernels which adds convenience without forcing repackaging
> for every kernel.
> 
> The project home is at http://linux.dell.com/dkms/.  I've even
> previously created a dkms-ified ivtv RPM which uses DKMS.  That's at
> http://www.lerhaupt.com/mythtv/permalink/ivtv-0.1.9-25dkms.noarch.rpm. 
> The source RPM is there too.

I've also placed dkms in the ATrpms repo for easing installation.

Whether dkms is used or not, the resulting kernel modules should be
packaged and injected into the rpm database (one can argue about that,
the core argument in favour of doing so is ensuring intergability of
the binary bits).

These packaged bits would also require a name :)

The dkms/rpm modell I have been discussing with Gary (or more or less
trying to convince him to extend dkms to) is

A) Common src.rpms as known carring source tarballs etc.
B) foo-dkms-....noarch.rpm a repo maintainer creates. There are module
   sources prepared/embedded in dkms' framework.
C) foo-kmdls or kernel-module-foos that both repo maintainers and
   users can easily create out of B)

The benefits of dkms are that users can easily rebuild kernel modules
for new kernels (if they are faster than the repo maintainer ;) or for
custom kernels.

I would suggest to discuss pros and cons of dkms separately from the
foo-kmdl vs kernel-module-foo issue, because I believe these two are
orthogonal to each other.
-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.atrpms.net/pipermail/atrpms-devel/attachments/20040512/9771fae1/attachment-0001.bin