[ OK, Jason wins for "coolest toys"  :-]

I think we've lost focus on this thread.  I missed the start of
it, but it's now focused on micro- vs. "algorithmic" optimization.
I don't think anyone's arguing that micro-optimization doesn't
(usually) give you the kinds of gains in performance that
improving the algorithm can.

However, once you've improved the algorithms, micro-optimization can
help performance of real-world applications--some applications more
than others, certainly, but even I/O bound applications can gain a
small amount from an optimized syscall interface.  What's wrong with
that, if someone has the time, energy, and inclination to make those
optimizations?

I think it's great that David and others have the time to spend on it.
That's one of the great things about free software--they're free to do
so, and it's great that Linux gives them the framework to do it in.
I think it would be neat if there were some NetBSD and Linux folks who
could engage in -*_friendly_*- competition to try and out-do each other
in performance and features on as many architectures as possible.

What I've seen is that the usual gambit seems to be trying to out-do
each other in denegrating the other which is a useless past-time unless
you feed off negative vibes.  I don't.

NetBSD could certainly use both kinds of optimization.

-allen

-- 
              Allen Briggs - end killing - briggs@macbsd.com