>The DECStation responds to pings directed at 224.0.0.1, so multicast must
>be working at some level...  Perhaps I need a new xntpd?

Uh, as of the release we ship, auth defaults to "on". Have you set a
key with type "M" (md5) ?


Have you precomputed the multicast/broadcast delay?


>The clock on the DECStation really shows how badly PC clocks drift.  It's
>rediculous.  :)

(cough) thank you.  


The DECstation timekeeping code has had, perhaps, the most attention
of any NetBSD port -- it has both Dave Mills' code and my own tweaks.
On a /240, it should hover around 10ppm off a good refclock, due to
the divide approximation and vagaries in the TC bus-clock crystal.  In
a temperature-controlled environment, it should be very, _very_
stable.  Or at least the /240 I had in a lab was.

(comparison between my 5000/150 and Michael Hitch's Personal
Decstation suggests the motherboard clock has lower jitter than the
r4000 CPU cycle-counter, so we only use the CPU cycle-counter on the
5000/150, which doesn't have any high-res time source on the
motherboard. I'm interested in talking to would-be time-vultures about
any improvements over the current state of affairs.)