[ntp:hackers] Re: Clock selection?

Frederick Bruckman fredb at immanent.net
Mon Jan 19 12:44:21 PST 2004


[I hope you don't object to the change in venue. Your comment in the
newsgroup triggered this thought I've been dwelling on for a while.]

In article <400ABDE2.F767B4D9 at udel.edu>,
	"David L. Mills" <mills at udel.edu> writes:
> 
> Ordinarily, delay is not a factor in the algorithm, as satellites can
> make very good sources. The usual problem is not the satellite itself,
> but asymmetric paths, one way by satellite the other by landline.

"Ordinarily", eh? I'm finding that it's very helpful to filter
out packets with large delays for a couple special cases. For the
intermittently connected dial-up, it rejects the packets of 1 to 3
second delays that were waiting in the queue when the link came up,
which makes a huge difference in stability. For the congested router,
it's also clearly helpful, though not quite as dramatically.

Tweaking the limit in "ntp.h" is not going to be acceptable for the
general case, but couldn't we have an option in "ntp.conf"? I have
already experimented with adding it to either of "tinker" and "tos",
but I think now it doesn't really belong in either one. So where
should it go? Should "maxdelay" be a top-level keyword, or should it
go under a new top-level keyword, and what would the new one be?

Frederick



More information about the hackers mailing list