Uh, Jonathan...

This was Jason, I think:
> >As far as I'm concerned, it doens't prove anything, because the
> >hardware was not identical.  I don't particularly care that the
> >hardware in machine A is "slower".

Thats you:

> That's your prerogative.  dont be surprised if the rest of
> the world thinks it does show something.

Uhm.

I've seen overall "faster" machines be slower in certain contexts, especially 
interupt latency.

E.g., if your disk drivers naively use the flush all data cache
operation on 68060/50MHz machines, maximum interupt latency goes
high enough that you can't keep up with more than 38400 bps at the
intermal Amiga serial port. Although the 68060 is a much faster
cpu than the 68040/40MHz or 68030/25MHz (I've seen especially the
latter keep up with much higher bitrates). (Where the cache
is smaller; and we're talking about 8+8kB vs. 4+4kB vs 256+256 bytes here).

This is not only theoretical; people have FIRST reported this and i did
the calculations later. 

Reading through some A. Brown /M. Seltzer papers, I see much stranger (to me,
who doesn't know that CPU very well) effects described in different
i586 boards, depending on what kind of bus cycles are created for
what kind of operations etc. Not all of them had a software workaround as easy
as the above described 68k effect has.

So, as long as you don't tell me what you did to exclude such effects (well,
that you considered, and excluded them), or, don't tell me you used
identical hardware (better: the same; best: two or three different
machines; each running both tests) for both tests, I'm not convinced.

> But if you _really_ want to do a head-to-head comparison, perhaps the
> most convicing way is to stop crapping all over non-PPS support and
> actually get PPS support in the NetBSD kernel.
> Then we could *do* the head-to-head test, very simply:
> just run NTP on the machine on top of both FreeBSD and NetBSD.
> 
> Does that really have to be spelled out explicitly?

Until this message of you, I didn't understand what you were aiming at.

Regards,
	-is