>Theodore Y. Ts'o wrote:
>> Quite frankly, at this point, the only driver which I really trust to be
>> well implemented and whose hardware is stable is the BusLogic series of
>> drivers....
>This brings up something I was mentioned a while back without any
>takers. One of the reasons Solaris is so stable, especially under
~~~~~~~~~~~~~~~~~~~~
>loaded conditions, is that Sun tests the hell out of their OS.
You've probably been using another Solaris than I do. Developing
Software I've regulary seen Solaris boxen go up in smoke. Most
important, don't let your software shout at the serial lines, and
don't trust the Solaris NFS client. (The NFS server seems fine, though)
>Wouldn't it make sense to have a test suite for the Linux kernel? It's
>pretty sad when Linux has been in production machines for years, and the
>above is the confidence level people hold Linux and its drivers up to.
>A test suite to stress-test as many scenarios as possible would be nice,
>and would help establish a minimum standard of performance/stabliity for
>drivers in the kernel.
Since the kernel FAQ <URL:http://kernelfaq.iconsult.com> I started
already contains a simple SCSI test, I'd be happy to collect more and
more tests until we have a kind of kernel test suite.
Froh
P.S.: The first 'official release' of the kernel FAQ will be today.
-- Frohwalt EgererVI has been created to tempt the faithless. C-u 1 0 0 0 M-x praise-emacs