Newsgroups: comp.protocols.time.ntp
Path: utzoo!utgpu!watserv1!maytag!watdragon!spurge!ccplumb
From: ccplumb@spurge.uwaterloo.ca (Colin Plumb)
Subject: Re: NTP and leap seconds
Message-ID: <1990Nov5.170725.20751@watdragon.waterloo.edu>
Sender: daemon@watdragon.waterloo.edu (Owner of Many System Processes)
Organization: University of Waterloo
References: <901105.134958.+0100.af@sei.ucl.ac.be>
Distribution: inet
Date: Mon, 5 Nov 90 17:07:25 GMT
Lines: 46

In article <901105.134958.+0100.af@sei.ucl.ac.be> af%sei.ucl.ac.be@CUNYVM.CUNY.EDU ("Alain FONTAINE ", Postmaster - NAD) writes:
>Big hi to all the nanoseconds watchers. I am quietly following this discussion
>since quite a long time, only to regret that the holy scriptures are not
>available to me (cannot print Postscript).

There are versions (minus all the figures) of rfc1119, rfc1128 and rfc1129
on louie.udel.edu in pub/ntp.  The rfc1119 is actually somewhat new & improved
over the Postscript version.

> Now finally I have seen a little
> bit of it translated for poor ascii suckers. And it raises questions :
>
>- the Julian Day Number of Jan 1, 1 is given as 1721426; both the program
>  I use to compute such things and a few solid astronomical tables give
>  1721424.

One thing to watch out for here is Julian vs. Gregorian calendars.  From
Calendrical Calculations (latest Software Practice & Experience - neat
article), extrapolating the Gregorian calendar backwards to Jan 1, 1 AD
gives a monday, which is Jan 3, 1 AD in the Julian calendar in effect at
the time.  So it may be that the Julian day number is correct using
extrapolated Gregorian (which it is defined to use), but two days off
if your tables use Julian dates way back then.

>- the MJD of Jan 1, 1972 is given as 15021 ; on the other hand, both the
>  program and the tables give 2415021 for the JDN (at noon), which means
>  2415020.5 at 0h and, substracting 2400000.5, 15020 for the MJD.

Er... 15021 is given for 1900; 41318 is given for 1972.  I dunno about 
the tables you have, but the NTP tables don't quite agree with common
sense.  There are an even number of days (366) between 1 Jan 1972 and 1 Jan
1973, yet the NTP chart shows MJDs of 41318 and 41683, which, you will notice,
are 365 days apart.  I figure the 1 July 1972 day (41,500) as being correct
relative to 1 Jan, but somewhere in July through December, a day got
lost.  (The dates in the table are mostly the day before.)

>- the values listed for the Dec 31 of various years are correct, given the
>  fact that one speaks about 24h, and not 0h.

In that case, the others are off.

>- but what about the various Jun 31 ?????                      /AF

No idea.
-- 
	-Colin
