>
> What's with the rush to push 2.0.0 out the door? It seems (from the
> number of Oops reports) that 1.3.9x and 1.99.x are far from stable. I've
> got 1.99.1 #1-pre-2.0 running on a test box (old a.out system) and its
> gone 8 hours with very little load (other than a few runs of make -j
> zImage :)...but 8 hours doesn't make a kernel "stable". Just last night,
> a system I thought was stable (1.2.13/1.2.14 cyclades terminal server) died
> horribly after 15 days.
>
There seems to be the mythos of a "stable kernel" taht works like
this: All the 1.3 kernels are terribly instable, even 1.99.99, but
1.2.x is always stable. Once we've gone from 1.99.99 to 2.0 it's a
stable kernel.
That's not how things are. Hey, 1.2.13 has real bugs! And the number
of bug reports also isn't a measure of the number of problems (just
remember the "ps -u dies on floating point exception" and broken
flock()". I think it's from Kernighan and Plaugher who said that "A
bug that always appears is an easy one; the bug that most of the time
does not appear is the hard one". With that in mind the recent oopses
were "easy ones". With minimum quality control from Linus, using 1.3
kernels shouldn't be a problem (if you don't control a nuclear
plant).
One of the problems with 2.0 is that Linux wants it to be stuffed
with "hering", and thus the usual release cycle (one per year) is
overdue. The code freeze was too late, but don't we like all the new
features? We don't need to hurry with 2.0, but we should really try
to get there soon!
Ulrich