In message <20050223072258.GM4454@snowdrop.l8s.co.uk>,
David Laight writes:

>Because I am thinking of server data flows - where there are multiple
>applications generating ther data.  

Then you're looking at the wrong problem. The problem under discussion
is a single very large flow (about 4 times what the fastest-available
single CPU for NetBSD can achieve today now).  If we can handle that,
then handling multiple modest flows is (by comparison) trivial.

I also suspect you hare fallen victim to the fallacy of implicitly
assuming that "server" machines are necessarily faster and
better-performing than any given "client" machine. That may be true in
many common scenarios, but it's not true in general.

I could repeat a couple of observations on near-term CPU performance
growth and on 10GbE price curves that (for example) Steven Bellovin
and I made here and elsewhere over the last month. The key points ---
an industry shift from exponential single-CPU performance growth to
multi-core CPUs, coupled with a 10x increase in speed of affordable
(but high-end) network interfaces -- is very real, easily verified,
and of great interest to some of us.  If it's not of interest to you,
okay, but please don't obstruct those who *do* want a solution for it.