RE: Fwd: Proposed Simple Text Format

["EXT Carl Stephen (Steve) Smyth" <stevesmy@microsoft.com>]

Hi Patrik

I hadn't thought about this in enough detail.

You're right in two senses. The fractional parts of the difference between
GPS time and UTC can vary up to a microsecond or so before the MC is
corrected. One part in a million doesn't seem like much, but could be
important in operations where some information was referenced to GPS time
and some was referenced to UTC. It also makes sense to hide the complexity
of any correction (using information in the NAV message or other source)
from network applications. So UTC is the right choice.

I found reverse engineering my faulty thinking interesting...in many
applications a small and bounded time difference of less than a microsecond
can be ignored and the statement that UTC - GPS_time = <net number of leap
seconds> is true. In my usual range of applications (personal and in-car
navigation), I never need to have a correspondence better than about 5 000
microseconds between the two since all positioning calculations we do use
GPS time as a reference. But I see your point when I broaden my thinking (as
I probably should have done sooner).

...steve


-----Original Message-----
From: Patrik Fältström [mailto:paf@swip.net]
Sent: Saturday, July 01, 2000 1:37 AM
To: Carl Stephen (Steve) Smyth; 'EXT Tim Salo';
ietf-spatial@research.nokia.com
Subject: RE: Fwd: Proposed Simple Text Format


At 08.53 -0700 00-06-30, EXT Carl Stephen (Steve) Smyth wrote:
>Actually GPS time started out as UTC and simply omits the leap seconds.
Leap
>seconds may be added or subtracted to UTC(only added to UTC so far) but the
>relationship between GPS time and UTC is well defined and always consists
of
>an integral number of seconds. Using either in conjunction with position
>information is fine as long as the time is identified as UTC or GPS time.

Actually, no. There is a difference which is also fractional. The 
master time in the GPS satellites is adjusted now and then to stay 
within one microsecond of UTC, and for things that uses the GPS time, 
there is a specific adjustment one have to do to get accurate UTC.

See ftp://tycho.usno.navy.mil/pub/gps/ser42.dat for the current table 
of difference between GPS and UTC.

One nanosecond error in the GPS is one foot in distance error...

My only message was that GPS is not accurate enough for people 
working with time  -- and especially the ones which try to build NTP 
servers with low stratum.

The time in these messages should be like in most other protocols on 
the Internet be UTC and not GPS (and I am not only talking about 
integer second corrections).

     Patrik