Re: Linux is moving too fast

On Tue, 23 Oct 2007, Damien Hull wrote:

> I only read the first article.
>
> As a non programmer I don't think I can comment on kernel development.
> Unless I get a broken kernel. In that case I would be complaining. This
> leads me to my first question.
>
> How often do Linux distributions such as Red Hat or Debian get released
> with a broken kernel?
>
> The reason I ask this question is because Linux "users" get their kernel
> from the distribution maintainers. I get my kernel from the Ubuntu
> developers. So far I haven't had any problems. I haven't seen any one
> else complaining either.
>
> Are the Distribution maintainers such as Red Hat or Debian complaining
> about the kernel and kernel development?

I can't say with any surety, but not long ago all the major commercial distros
(with the possible exception of Slack) were shipping heavily patched
kernels. They'd typically lag behind the latest kernels but back-port the
fixes and features they wanted. That practice tended to insulate them
somewhat from the flakiness of interim releases.

That might not constitute vocal complaining, but it's certainly an implicit
and tacit admission of the issue.

> Now on to something a little more fun. Well, maybe.
>
> 1. Do you compile your own kernel?
> 2. If yes why?

I do, and for more than just because I maintain a distribution. I do so
because even with the majority of support compiled in as modules it bloats
the kernel size. Fat kernels are less cache coherent.

I also do it because auto-probing device support can be problematic when
there's more than one driver for a given device. There's legacy and newer
drivers for several SCSI HBAs and NICs, for instance. The wrong driver can
cause performance and/or stability problems. Worse, some drivers won't work
at all when loaded with other drivers. The Fusion MPT drivers, for example,
wouldn't work in a kernel that also had I2O (IIRC) support loaded during the
early 2.6 days.

I've also had to maintain my own out-of-tree patches to fix things like BSD
process accounting for 32-bit UID/GIDs. It literally took years for the
kernel maintainers to agree to a final fix for that one.

I also have a half dozen reasons that are more personal idiosyncrasies and
philosophical, so I won't go into that. Suffice it to say that there are
very good reasons why you'd want to compile your own kernel.

         --Arthur Corliss
           Live Free or Die
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Wed Oct 24 10:36:04 2007