Re: What your CIO is reading ....

On Thu, 11 Oct 2007, Mike Tibor wrote:

> I actually haven't had any problems on any of our boxes which are running
> 2.6.x kernels. What specific problems have you had? Could they be
> related to pSeries-specific code?

Bare in mind that pSeries features is one reason why I've moved to 2.6, but
I've had issues on x86 platforms as well. Since I have my own distribution
to maintain it's just simpler to run the same kernel rev across all
architectures.

In the 2.6 series at various times they've broken the RAID5 driver (during
the merge w/RAID6 driver), multicast support on virtual ethernet (ppc64),
the minix filesystem (64b everywhere -- and yes, I did use minix for initrd
until that point), and combinations of iptables modules (like mixing
multiport and ip range modules), but the iptables bugs may have been present
in the 2.4 codebase as well. There's been more than that, but that's off
the top of my head.

One real frustration, as well, deals with migrating to newer kernel revs. I
can't tell you how many times I've upgraded my kernel when building a new
distribution build, only to find out I'm missing half of my standard device
and feature support. Everytime they reorganize the kernel feature tree or
rename the features it breaks 'make oldconfig'. It's happened frequently
enough that I know do a tedious manual menuconfig after oldconfig and verify
every item I standardize on. That only happened one time during the 2.4
series.

Look at all the fun they've had reworking major portions of the schedulers
and memory management. They've broken a lot of out-of-tree drivers. The
kernel module API alone was a hell of a lot more stable before 2.6. If they
want to make major disruptive changes that should happen in the development
branch (what 2.7 should have been), and just integrate bug fixes into 2.6.x.

I could go on (believe me, I'm only getting started), but I won't. Suffice
it to say that my life as a maintainer was much simpler in the pre-2.6 days.

         --Arthur Corliss
           Live Free or Die
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Thu Oct 11 17:58:23 2007