kernel modules/kmods/dkms (Re: Plan for tomorrows (20070906) FESCO meeting)

[Thorsten Leemhuis]

On 05.09.2007 19:01, Rahul Sundaram wrote:
> Brian Pepple wrote:
>> Hi,
>>
>> Please find below the list of topics that are likely to come up in the
>> next FESCo meeting that is scheduled for tomorrow, Thursday at 17:00 UTC
>> in #fedora-meeting on irc.freenode.org:
>>
>> /topic FESCO-Meeting -- MISC -- obsoleting kmod proposal:
>> http://fedoraproject.org/wiki/DavidWoodhouse/KmodProposal - dwmw2, f13 
> 
> Is the alternative proposal using DKMS at 
> http://fedoraproject.org/wiki/JefSpaleta/DKMSProposal being discussed too?

Well, I think lots of the reasons against Kmod's are similar for dkms
(or any other packaging standard for kernel modules); quoting some part
from http://fedoraproject.org/wiki/DavidWoodhouse/KmodProposal
below:

> There is no justification for shipping kernel modules as separate packages within Fedora. If code is good enough to ship, it should be shipped in the kernel packageIf it's not, then it should not be shipped at all with the 'Fedora' name on it. 

Same for kernel modules packages of all kind.

> Kernel module packages just cause gratuitous complexity for RPM/YUM and for users, without any real benefit.

dynamically building modules on users is IMHO also "gratuitous
complexity". One of the reasons: modules now and then simply won't
compile anymore when the kernel-api changes; thus dkms tries to rebuild
the modules on lots of systems and will fail on lots of systems.

> Where there are certain drivers which are almost ready to be merged upstream and which Fedora would benefit from shipping, we have always managed to do that. [...]

Same for dkms

> Instead of shipping kmod packages, we should consider adding patches to the kernel RPM instead. [...]

Same for dkms

> The complexity of separately-packaged kernel modules is unnecessary, and the users' problems with upgrading when the modules are not built synchronously with the kernel will no longer be possible.

Same for dkms, as modules might break if the api changed.
--

Further: the rules for including kmods in Fedora were and are quite
strict. They were the main reason why we only had two kmod's (which was
one of the reasons why we stopped developing kmod's further). So if
those rules won't change (which I doubt) it's IMHO not worth investing
work in another ruleset for kernel modules in Fedora.

Just my 2 cent.

CU
knurd

P.S.: No, I don't have anything against dkms in particular -- dkms or a
dkms-like solution in combination with pre-compiled module packages
(/like kmods or kmdls) are IMHO what we should aim for. But I agree with
dwmw2 that Fedoras proper repos are not the place for it anymore. It was
something different when we had Core and Extras separated.