[ntp:questions] Re: ntpd transmit timestamp precision

Harlan Stenn stenn at ntp.isc.org
Mon Feb 20 03:35:16 UTC 2006


David> I'm uneasy calling random() for every read of the clock, as
David> it could be expensive.

We could fairly easily fuzz 1 out of every N calls or fuzz a random
percentage of calls (we could use the value of ntp_random() for this purpose
as well as for the fuzz).

David> It seems a bit overkill to fuzz the nanobits
David> when the caller has timespec and fuzzed to 10 ms. Perhaps a test of
David> precision greater than a microsecond is advised.

That's what my patch would do - we could easily adjust the threshhold, or
even offer a tinker interface.

Want me to code it up?


More information about the questions mailing list