John Hasler wrote:
Richard B. Gilbert wrote:
I suspect that faster convergence would carry a price such as large
amplitude oscillation!
Chrony converges faster and doesn't oscillate. I would like to see a
rigorous analysis of the algorithms used by Chrony, though.
If you can't stand it, roll your own!
Richard Curnow did and we now have Chrony.
You may find that it's far more difficult than you think!
It wasn't easy.
I really don't understand why Rob will neither switch to Chrony nor fork
Ntpd, since he is clearly dissatisfied and frustrated by Ntpd as it
stands. Both these choices are open to him, as is writing his own
implementation from scratch.
Seems another possible alternative is on the way
http://www.cubinlab.ee.unimelb.edu.au/radclock/
At moment it's Linux/FreeBSD only, I guess it will be same as
for recent versions of ntpd and chrony and not be usable on
NetBSD.
David
_______________________________________________
questions mailing list
questions@lists.ntp.org
http://lists.ntp.org/listinfo/questions
