 |
 |
>> If people follow the ntp docs, they should NEVER trust any single >> source for reliable time. Right, but for a long time I've been polling two servers (ntp.astro and madhaus.utcs). What I don't understand is the behaviour of the ntpdate command if two servers are queried, and the times reported differ by a lot: It turns out that all of the following ... ntpdate 128.100.77.35 ntpdate 128.100.102.10 128.100.77.35 ntpdate 128.100.77.35 128.100.102.10 ... cause my system's date to be reset to the bogus time from ntp.astro! Not an average of the two, not the ntp.astro one with a warning about a large mismatch between the two. Just the ntp.astro one, as if that were the correct thing to do. The ntpdate manpage doesn't provide a lot of help in terms of telling me what algorithm it uses to pick the "right" time if two servers are used, and they don't agree. It does say, though, that better performance and greater resistance to insanity on the part of any one server will be obtained by providing at least three or four servers, if not more. So this brings up the obvious question: Besides the seemingly-reliable madhaus.utcs, and the perhaps-someday-reliable ntp.astro, is there some third system on campus that it would make sense to add to our lists of NTP servers to poll?