In message <20030824234822.CECE890@coconut.itojun.org>Jun-ichiro itojun Hagino 
writes

>	then all 4.4BSD-based systems (i.e. without hash lookup) are toy-system

In today's environment, for deployments with thousands to 10,000
simultaneous connections on a single machine: sure, CSRG system
without a hash-table would get the same label, at least from me.
That *was* nearly ten years ago.


>	i don't think hashed inpcb lookup alone can classify operating system
>	between toy-system and non-toy-system.  i don't see why you bother
>	to make a comment like this.

And I don't see any grounds why any reasonable person to object to it.
I qualified my statement quite carefully.  For the class of
deployments I described, what I wrote isn't even particuarly
perjorative.  It is a plain, straighforward statement of the facts.

See again what I said about thousands simultaneous connections, and
what I wrote earlier about handling around 200,000 packets/sec.  Do
you think a TCP with no hash table is a reasonable (or even a
justifiable) choice for that environment?

I don't.

And I am genuinely quite surprised to learn that NetBSD (nor, I
gather, FreeBSD) are, by those standards, adequate *IPv6* platforms
for that kind of workload.  Until today, I had blithely assumed the
(Net)BSD IPv6 was of comparable quality to our IPv4 code.