NavList:

In addition to the round-trip travel time issue that has been mentioned, which is the most important trick, there are at least two other factors. You mentioned "vagaries of internet transmission times". Actually, the internet is really fast, but we often still think of it as fairly unreliable because we're used to web pages as our primary interaction. Web pages are loaded using a transport technology that does not tolerate errors (of transmission), and as a result it's quite a bit slower than the actual internet speed. An alternative is a transport technology called UDP which is also used for things like VoIP phone calls. It is very fast, but has no error checking which doesn't matter for very short data messages like getting the time from the NIST site. Another issue, in this particular case, is the local implementation of Java on your machine. I checked the estimated error in the NIST time display using various browsers and different machines and operating systems. A Mac running the lastest OS X and the standard Safari browser had the lowest estimated errors, consistently 0.1 seconds, while a Windows Vista machine using a variety of browsers had errors usually between 0.2 and 0.4 seconds. You can do some sanity checks on these things by opening several browser windows of the same clock site. You can watch them all updating on screen and differences of a few tenths of a second are quite noticeable.

-FER

----------------------------------------------------------------
NavList message boards and member settings: www.fer3.com/NavList
Members may optionally receive posts by email.
To cancel email delivery, send a message to NoMail[at]fer3.com
----------------------------------------------------------------