Posted by Moe Trin on 07/03/07 01:45

On Mon, 02 Jul 2007, in the Usenet newsgroup comp.os.linux, in article
<NkYhi.5342$Io4.64@edtnps89>, Unruh wrote:

>ibuprofin@painkiller.example.tld (Moe Trin) writes:

>>Maybe you want to read the NTP documentation. If all you have is one
>>time source, you are indeed unable to correct for propagation delays.
>>But if you have more than that (can you say "redundancy" Bill? I
>>though you could), you can deduce and eliminate nearly all of the
>>delays.
>
>What delays? The delay between the gps receiver and the computer it is
>attached to is a usec at most.

Geez Bill, don't quit your day job. You are assuming that the delay
between the server application producing the time stamp, putting that
time stamp into a UDP packet, getting that packet out onto the wire,
up through the network stack on the receiving end, and into the NTP
client a being zero, and only the transit time of an IP datagram over
the few meters of local wire matters. Wrong-O By the same token,
the GPS satellites are not located just down the hall but are more than
a few miles above the surface of the geoid (try "thousands of miles"),
and are changing the absolute range at a fairly significant rate. The
circuitry in the GPS receiver needs the signals from _four_ satellites
at the same time in order to determine absolute position of the
receiver, has to listen to the ephemeral data contained within the RF
bitstream in order to see where the sending satellites are, AND THEN
make a correction for the range between each satellite and receiver.

>However redundancy is good, esp is someone decides to disconnect your gps.

They're on UPS, just like the rest of the servers. Doesn't prevent the
power from being disconnected, but as they are in a secured server room,
it reduces the possibility some.

>I guess you meant that you have one gps receiver which acts as a stratum
>one to a whole bunch of other systems. Then yes, redundancy can be useful.

Didn't take the hint? Let's try again:

-rw-rw-r-- 1 gferg ldp 43295 Nov 18 2005 TimePrecision-HOWTO

>On the other hand just put one on each critical machine and be done with
>it.

A bit impractical in these building which by reason of the metal
fabrication, insulation with vapor barrier, and distance from a location
where you can install a GPS antenna makes it a pain to do so. Oh, and don't
forget to apply a correction for the length of the cable between the
receiving antenna and the actual location of the receiver.

>>Read the docs from Dave Mills at UofDelaware. He writes pretty well.
>
>Yes, have done so. Just which point was it you wanted me to read?

[compton ~]$ zgrep -c propag rfcs/rfc4330.txt.gz
5
[compton ~]$

There might be a good word to look for and one of several places to start.

Old guy

[Back to original message]