[plug] Linux fails the high-end test

[Greg Mildenhall]

On Wed, 21 Apr 1999, John Summerfield wrote:
> > On Wed, 21 Apr 1999, Bret Busby wrote:
> > > I have not read the printed issue of the Australian, and the story is
> > > not in the online published set of stories.
> > Its actually a good article. Pity about the headline, as is often true.
> I've seen claims that the Linux wasn't set up very well on the test 
> system, that the test was paid-for by the Evil Empire...

Those tests were for a different report. (by Mindcraft) The DHBrown report
had no empirical data whatsoever, and was more about what GNU/Linux was
being used for than what it is capable of being used for.

> One thing I'm sure of: if the test wasn't done with a 2.2 kernel, it's not 
> worth the paper it's written on.
The Mindcraft benchmarks were done on a 2.2 kernel, but a version that was
2 revisions old and known to trigger a flaw in Window's TCP/IP stack,
resulting in poor performance.

> Some other information wasn;t entirely accurate: 2 Gb RAM may be the limit 
> on Intel (I simply don't know), but on 64-bit platforms?
There were dozens of misconfigurations involved. Some of them were perhaps
just carelessness or lack of research, but one or two of them appear to be
deliberate changes made to lower performance.

-Greg