Anthony DeRobertis wrote:
On Sat, Nov 10, 2007 at 10:41:39AM +0100, Rob Janssen wrote:
You could have big trouble when using a satellite-based access
(especially with some other means of uplink), but fortunately that is
not common.
I've used ntpd over two-way satellite. It worked (though of course I'm
sure the accuracy was degraded). 600ms RTT was no fun.
Well, with a commercial satellite link with resonable capacity planning
that will probably work.
But with those "broadband via satellite for consumers" deals it will be
no fun... those a so badly overbooked that the consistency of the
roundtriptime is awful. Here are some results of pinging one such system:
PING victim (victim) 56(84) bytes of data.
64 bytes from victim: icmp_seq=1 ttl=47 time=1166 ms
64 bytes from victim: icmp_seq=2 ttl=47 time=1229 ms
64 bytes from victim: icmp_seq=3 ttl=47 time=1180 ms
64 bytes from victim: icmp_seq=4 ttl=47 time=1020 ms
64 bytes from victim: icmp_seq=5 ttl=47 time=1321 ms
64 bytes from victim: icmp_seq=6 ttl=47 time=1172 ms
64 bytes from victim: icmp_seq=7 ttl=47 time=905 ms
64 bytes from victim: icmp_seq=8 ttl=47 time=1062 ms
64 bytes from victim: icmp_seq=9 ttl=47 time=1150 ms
64 bytes from victim: icmp_seq=10 ttl=47 time=1652 ms
64 bytes from victim: icmp_seq=11 ttl=47 time=2096 ms
64 bytes from victim: icmp_seq=12 ttl=47 time=1611 ms
64 bytes from victim: icmp_seq=13 ttl=47 time=1214 ms
64 bytes from victim: icmp_seq=14 ttl=47 time=880 ms
64 bytes from victim: icmp_seq=15 ttl=47 time=1210 ms
64 bytes from victim: icmp_seq=16 ttl=47 time=1178 ms
64 bytes from victim: icmp_seq=17 ttl=47 time=1116 ms
64 bytes from victim: icmp_seq=18 ttl=47 time=952 ms
even within these 18 samples, the time varies like 1200ms. no way any
accurate time referencing can be done.
probably ntpd will not even lock correctly.
in this case it could be a better idea to lock to the time-of-day
messages on the transponder (after verifying they are somehow set to an
accurate time standard)
Rob
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
