RE: No Escape - Internet-based atomic clocks

["Albertson, Chris" <CAlbertson@primeadvantage.com>]

You phrase "do not always" is ezactly dead on.

This implies that "Some times they do work".  The one
program I use and trust is that same exact program used
by _all_ of those authoratative source out there on the
Internet.  Kind of make sense to use what they use.
This is called "xntpd" and was developed by the person
who also developed the NTP protocol.  Dave Mills.

>From a technical standpoint this is a very interresting 
algorithum.  It does something most people would guess can't
be done.  It's a clasic servo control problem:  Keep a
local clock servoed to a "master" clock. The interresting
part is that the clocks stay in sync orders of magnitude
better then the uncertainty in the communications delay.
I would have guessed that if there has a 100ms one sigma
in the comm delay then the local clock would have a 100ms
one sigma error.


> -----Original Message-----
> From: David Troyer [mailto:dst@uni-muenster.de]
> Sent: Wednesday, March 14, 2001 2:45 AM
> To: 'Tass Mailing List'
> Subject: RE: No Escape - Internet-based atomic clocks
> 
> 
> I have read that Internet-based clocks do not always 
> synchronize the PC 
> exactly with the atomic clock, because of lags incurred 
> during time-serving 
> from the standard atomic clock through the Internet itself. 
> As I recall, 
> the time error is variable and of the order of zero to 
> several hundred 
> milliseconds, so one may have to be careful if a high degree 
> of accuracy is 
> required. Chris Albertson has also discussed the perils of 
> SNPT vs NPT (two 
> mails, RE: No Escape).
> 
> However, if these objections are not of consideration, then 
> Tom could try 
> 'AtomTime' instead of Tardis. AtomTime is freeware, only 32 
> kb in size and 
> the source is perfectly documented, so the executable can be 
> tailored as 
> desired.
> 
> Best wishes,
> David
>