Re: [LEAPSECS] how do computer people want their time clocked?

From: Deckers, Michael <Michael.Deckers_at_fujitsu-siemens.com>
Date: Fri, 01 Jun 2001 14:49:36 +0200

   On 2001-05-31, Paul Eggert wrote thusly:
>
> ... ITU-R TF.460-5 (the current standard for UTC) .............
>
   Right, I should not have quoted the text from [ITU-R 460-4].
>
> The body of the standard refers to UTC as a count of seconds in
> several places: it talks about UT1-UTC, UT2-UTC, and UTC+DUT1, and all
> of these expressions make sense only if UTC is a count of seconds.
> For example, see Annex 1.D, which defines DUT1 = UT1-UTC.
>
   Good point -- these "counts of seconds" are what astronomers
   call timescales. To phrase the issue at hand more clearly
   I probably should have asked:

      At which instant during the last leap second insertion
      did the value of DUT1 increase by 1 s, at the beginning
      of the inserted leap second, or at the end?

   If I understand you correctly, you say it happened at the end
   of the inserted leap second, together with the minute marker
   in the UTC signal.

   My assumption was that it happened at the beginning of the
   inserted leap second, so that UT1 - DUT1 was before
   1999-01-01T00:00:00 during the whole leap second, and so that
   the leap second can truly be said to be inserted before Jan 01.

   I admit that my reasoning is weak (and I also admit that
   DUT1 is only an approximation to UT1 - UTC, so that DUT1 could
   have changed by 0.9 s or 1.1 s and the argument above is not
   quite correct).

>
> > I fully agree that 1 d = 86 400 s, no matter which timescale.
> More than one ISO standard agrees with you. And yet this statement is
> incompatible with the common and natural interpretation of ITU-R
> TF.460-5, the interpretation that says that 1998-12-31 was exactly one
> day long at Greenwich.
>
   I do not see the incompatibility.
   If "the length of 1998-12-31 at Greenwich" is meant to be

      the difference between the readings "1998-12-31T00:00:00"
      and "1999-01-01T00:00:00"

   then that length is 1 d, no matter what timescale has been used.
   This looks like a "natural intepretation" to me. But if the
   "common and natural intepretation" is something like

      the difference between the readings of TAI at the instants
      at Greenwich when the readings of UTC were "1998-12-31T00:00:00"
      and "1999-01-01T00:00:00"

   then the answer may differ from 1 d, and you may even have to
   specify which instants you mean because instants with different
   TAI readings can have the same UTC reading. You are in effect
   comparing two different timescales -- it is like comparing distances
   by road and by rail. It is not surprising if the answer is not 1 d.

   Michael
Received on Fri Jun 01 2001 - 05:50:05 PDT