On Wed, 06 Aug 1997 01:33:02 -0700, I  <jonathan@dsg.stanford.edu> wrote:
[lots of stuff about tuning]

Hmm....

I should first have pointed out that what I'm upset about is a
difference that is a difference between 8-9 ppm and 55 ppm, or about
46 ppm. (or  46 parts in 2^20, to  be more precies).

If I can divide (not always a good assumption), that comes out at
about four CPU clock-cycles per real-time-clock tick.

To time vultures, that's still an order of magnitude, one which NTP
will detect, and which does cause xntpd to look down its nose at
NetBSD-based NTP servers.

As with the other performance issues... de gustibus non disputandum est,
or whatever.