On 12/1/06, Troy Dawson <[log in to unmask]> wrote:
> Hello,
> I have read up more on the kernel modules debate, as well as asked
> around some on private e-mail.  It's time now to bring the debate to the
> mailling list.
>
> As far as I can tell, there are three ways to do kernel modules.  They
> fall into three catagories.
> kmod - What Fedora and Redhat decided to do
> kmdl - What Axel Thimm has developed and implemented
> kernel-module - What we currently are doing, although I believe the idea
> was originally from livna
>

dkms is the fourth way and was being looked at for Fedora. At Sandia,
I had pretty much found that this was the only sane way to update 2000
desktops who might get kernels and special other things at any time by
scientists. The dkms had 2 nice things:

1) I could compile a one for the current kernel and it would get
picked up for the users.

2) Or if I didnt get it in time.. it would recompile itself when the
system booted into a new kernel, and tada happiness. When dealign with
engineers who like to add various tweaked things because they are
going to attach some wizbang unit they have.. it made our lives
easier.

dkms
Pros)
a. Even though it is dell branded, it isnt hardware specific and I had
them installed on various other hardware.
b. It knows how to upgrade itself which works 90% of the time. [I cant
remember where it didnt.. but putting in 10% just to satisfy Murphy]
c. It is maintained and advanced by Dell. It works with Fedora/*EL/SuSE.
d. Matthias Saou <[log in to unmask]> is using it for the rpmforge
items that he covers. Dag may also go to this.

Cons)
a. Its different from how you are currently using
b. It is yet another way of maintaining kernels.
c. You have to use a naming convention to say that dkms whatever means
a kernel module.


>   Troys Opinion:  It's obvious we need to either stay with kernel-module
> or switch over to kdml.  I vote for going with kdml, and make that into
> the standard.  It think eventually Fedora and RedHat will see that this
> way is less work, easier to implement, and just plain works.  I am not
> so set on this idea that I cannot be persuaded to continue with
> kernel-module.
>
> Hopefully this e-mail starts some good debate.  Please no flames.
> Everyone is entitled to their opinion.
> If you already responded to me in private, feel free to repost your
> opinions here.  I haven't ignored them.

Sorry for throwing this in at the 11th hour. I missed the whole
previous thread for some reason.


-- 
Stephen J Smoogen. -- CSIRT/Linux System Administrator
How far that little candle throws his beams! So shines a good deed
in a naughty world. = Shakespeare. "The Merchant of Venice"