USNO leap seconds - a minmum-change approach

From: Seeds, Glen <Glen.Seeds_at_Cognos.COM>
Date: Mon, 14 Aug 2000 10:15:17 -0400

Forgive my naive intrusion into this discussion. I have read the material on
the web site for this mailing list, and I don't understand why solving this
problem is so difficult.

The excellent article by McCarthy and Klepczynski seems to leave open the
possibility of a simple solution not explicitly stated. If I understand
correctly, the current system is in disfavour because of the technical
difficulty of re-synchronizing the world's UTC clocks at each change in a
way that does not disrupt time-sensitive systems. However, does this problem
not disappear if the change points are known precisely well in advance, and
the amount of change is known a little bit in advance?

For example, if the UTC change points are specified at a precise time in
June and December of every year, and the magnitude is known to be either
zero or one second a week in advance, then all systems can download the
magnitude prior to the change point, an apply it at the precisely specified
time.

Does this not resolve all objections?
- No laws need to be changed, because these dates have already been agreed
upon internationally as part of UTC.
- No synchronization is lost, because the change points are precisely
specified in advance.
- The "civilian" requirement that UTC and UT1 never differ by more than 0.9
seconds is preserved.
- Software change is minimal, because UTC already includes these change
points.
- This will still work at 2050, because up to two seconds per year can be
added.

Admittedly, this will no longer be adequate when there are more than two
seconds per year of incremental difference between UT1 and TAI. It would
also fail should the incremental difference become negative. However, this
interim solution should give us another 100 years to work that problem out.

What am I missing?
  /glen

Glen Seeds
Cognos, Inc.
1-613-738-1440
Received on Mon Aug 14 2000 - 07:26:47 PDT