2650 troubles, was Okay, this is so not cool..

Andrew Mann amann at mythicentertainment.com
Thu Oct 16 11:21:00 CDT 2003

Xose Vazquez Perez wrote:

> Marc Schmitt wrote:
>>Xose Vazquez Perez wrote:
>>>first mistake, 2.4.xx kernels of kernel.org are not greatly tested.
>>>It's better to stay with latest kernel of your distribution.
>>Huh? Elaborate on "2.4.xx kernels of kernel.org are not greatly tested"
>>please, I'm interested in how you could come up with such a statement
> here it goes.
	Can you back these claims up with specific versions and options of 
mainline kernels (stable/even series only please) that don't compile or 
crash while vendor provided kernels function fine with the same/similar 
build options?  In the case of security/critical bugs, you have to 
search pretty hard to find a kernel bug or this nature in the last 
couple years, but can you provide any examples of any bugs of this 
nature that were found and not fixed in a timely manner by a new kernel 
	It's always seemed to me that the specific vendor kernels have started 
with a kernel.org kernel, then added patches that were not accepted to 
the kernel.org kernels (for whatever reason, philosophical prominent 
among them I think), and then future patches always seem to be backports 
from the newest kernel.org kernels.  The exception being when the work 
for a bug fix comes primarily from one of the vendors, in which case the 
patch usually shows up in that one particular vendor's kernel a little 
bit before the kernel.org kernels, and in a little different form.

> kernels of kernel.org:
> - usually don't compile 100%
> - when is discovered a security or critical bug it is not released a new
>   formal kernel.
> - it passes no test, only "guy's boot tests"
> with these notes, some kernel gurus recommend stay with distributions kernels.
> This is not new.
> Distribution kernels, before to be released, pass heavy stress tests. In Red Hat
> Linux, at least one week.
	Wow, a whole week? :)

Andrew Mann
Systems Administrator
Mythic Entertainment

More information about the Linux-PowerEdge mailing list