[ntp:hackers] refclock_atom.c

David L. Mills mills at udel.edu
Mon Dec 20 19:48:03 PST 2004


Harlan,

Don't embrace this idea blindly. It can cause a lot of trouble under 
conditions where the delay is flapping largely in both directions, 
especially when clockhopping under those conditions. The minimum delay 
has large variations and causes even more jitter than without huffpuff. 
I learned this in early experimenting. The conservative and I think 
correct design is to link it with the prefer peer.

Dave

Harlan Stenn wrote:

>Mark,
>
>Please consider adding the per-{server/peer} huffpuff idea to
>
> http://ntp.isc.org/Dev/NewNtpConfFormat
>
>While Dave avoids that web, I'll at least be able to remember
>it that way.
>
>H
>  
>




More information about the hackers mailing list