Re: leap second policy making

From: Rob Seaman <>
Date: Fri, 21 Jul 2000 16:18:01 -0700 (MST)

I'll have to thank the time and frequency community for providing a
lot of good background reading material :-) I've read through the
McCarthy and Klepczynski article ("M&K") in GPS World (now that I know
it exists) and just finished the excellent "Greenwich Time and the
Longitude" by Derek Howse. The timing seems particularly appropriate
given the recent premier of "Longitude" on American TV. Harrison's clock
was also on the cover of the millennial Smithsonian magazine - quite a
happy choice for anybody who learned about Harrison's No. 4 clock from
Fr. Jenkins' History of Astronomy at Villanova a quarter century ago.

Folks are seriously underrating the legal ramifications of detaching
civil time from its terrestrial roots. I'd recommend Howse's book
for a historical perspective. For instance, the scientific community
expressed a variety of opinions regarding longitude and time at the
Seventh International Geodesic Conf. in 1883. Several of these were
overturned by the diplomats at the International Meridian Conference
in 1884. It is fundamentally the resolutions of that conference that
we are discussing changes to. These resolutions only achieved legal
status as each individual country adopted them in turn. Apparently
Liberia was a holdout as late as 1972.

This is not really a question of leap seconds - this is a question of
the international definition of the day. Even small changes to such
a fundamental legal (and philosophical) entity can be multiplied
immensely. Howse provides examples of litigation resulting in the
mid-19th century from the ambiguity between calendar dates as measured
by standard and local times (or mean versus apparent). By allowing UTC
and UT1 to diverge, a growing window for lawsuits opens in hundreds of
countries around the world. Farfetched? Ask a lawyer. Do we really
assert that litigation is less likely today than 150 years ago?

Note that currently UTC is tied to the location of the prime meridian.
One interpretation of a floating UTC is a floating prime meridian.
We may be generating a vast market for surveyors in Great Britain over
the next century as the meridian moves a couple of miles east of the
Airy Transit Circle. (Or is it west? Quick! Which way?)

The position of the international date line is still governed merely by
the conventions of nearby islands. Whether or not you regard the point
of view of a moving prime meridian seriously, nothing can stop the
islanders from doing so. This may not be a question of 140 seconds
(greater than the width of the Sun or Moon in the sky) over the next
century - for some this may be a question of a 24 hour discontinuity.

In any event, a new international definition of UTC would also very
definitely require a new international definition for the location of
the prime meridian.

M&K provide a good overview of the options facing civil time regarding
future leap second policy, but the motivation of the resulting survey
(and presumably, of the URSI Committee J working group?) for focusing
on only one of the authors' five options remains unclear.

Has there been a longer standing debate among the time and frequency
community about possibly discontinuing leap seconds? It would be very
difficult to find a single physical scientist on the planet to speak up
for M&K's option to redefine the length of the SI second to better match
the current length of the solar day (especially since this would also
only represent a temporary fix). But that leaves several options
available for tinkering with the scheduling of leap seconds short of
abandoning the community's entire obligation to reconcile terrestrial
time scales with atomic.

Perhaps this focus on one option represents a bit of a rush to identify
an answer now rather than a few decades or centuries hence when the more
rapid pace of leap seconds will only begin to intrinsically matter.
In the long term attention will certainly be required by the realities
of timekeeping - terrestrial and otherwise. But in the short term?

Imagine two scenarios. First, that the current status quo prevails (or
perhaps there are slight tweaks to the leap second scheduling algorithm).
Second, that the leap second is eradicated. Age the two scenarios by a
decade or two or three. Now then - decide that the original decision
was faulty. What happens? Under the first scenario the community (that
is, the world) simply implements whatever changes are needed at that point.
Perhaps we even move to the wacky "metric" time system of another book
I found in the library.

What happens under the second scenario? We have eliminated leap seconds
and allowed UTC and UT1 to diverge. Our children have decided it was a
poor idea. How do they fix it? Well - presumably they add back in the
time lag that has accumulated. If they are trying to fix the problem
it's because they decided the diverging time lag was too big. Now they
either have to swallow that lag in one big lump - or they have to issue
repeated leap seconds on an accelerated schedule. If one per year is an
issue - how about one per week or month?

M&K do offer a handful of reasons that action is required now. While
"a leap second does not affect GPS operations because its time system
is GPS time", we are told that GLONASS currently does have a problem.
But then we hear that "GLONASS developers plan to significantly reduce"
the problem with the next generation of satellites. (And are there
any estimates of the number of GPS users versus GLONASS?)

We do hear that "navigation is not the only service affected by leap
seconds" (although the above is the only - rather equivocal - navigation
example provided). The other "emerging problem" quoted is that "while
a leap second is being introduced, and until sychronization [sic] is
established, communication can be disrupted between some [spread-spectrum]
systems." Not all such systems apparently - just "some".

Are these indeed "emerging problems"? Or just temporary inconveniences
for some systems whose designs or implementations could be improved?
Is there any indication that the correct handling of leap seconds is
intrinsically impossible? After all, the title of the M&K article is
"GPS and Leap Seconds - Time to Change?" and we already learned after
just a few paragraphs that leap seconds have no affect on GPS operations.

Let me admit my woeful ignorance (not the only time) about other such
systems that might exhibit leap second handling problems - maybe these
are so well known in the community that M&K felt no need to mention them.

Perhaps readers afflicted by leap second troubles could detail these on
the mailing list?

Next M&K assure us that "significant variations [of the Earth's rotation]
prevent the prediction of leap seconds beyond a few months in advance".
The authors certainly should know about this - and this certainly appears
likely to be a continuing headache for the community. I presume, however,
that detailed studies of the Earth's rotation will continue for purely
scientific reasons - and that improvements in theory and computer
modeling will continue to be reflected in a deepening understanding of
such variations. Whether or not UTC continues to be slaved to UT1, the
relationship between the two will continue to be critical to many pursuits.

But this statement about the difficulty of predicting leap seconds for
civil time purposes is somehow connected by the authors to "the growing
urgency for a uniform time scale without discontinuities" for precise
time purposes. Later in the article, the authors recommend "should the
demand for TAI increase, timekeeping laboratories may need to consider
making this time scale more accessible to the user."

One may certainly believe that there is a growing need for quality time
products of many descriptions. But is the need for a precise, uniform,
non-discontinuous time scale being artificially confused with the
requirements of civil time - of UTC, that is - a time scale that is
required to be either discontinuous or non-uniform?

GPS already provides both types of time scales for many purposes. And
if a specific requirement is identified for improving the widespread
accessibility of TAI or related products (via GPS or otherwise) - what
does this really have to do with UTC?

M&K express that "concern is growing that users will construct time
scales independent of UTC that they perceive are more suited to their
individual requirements" and that "this would lead to an increased
number of nonstandard time scales". (To somebody on the outside of
this debate, it isn't even clear if GPS is considered to be a standard
or nonstandard time scale.) Perhaps the real point is simply that a
single time scale can't satisfy all users?

Take a larger view - any time-related project has requirements dependent
on the selection of an underlying clock. Is it realistic to attempt to
not only satisfy the needs of all current time and frequency projects
with a single clock, but also to predict the needs of all future such
projects? And even if a single clock is possible - is it the civil
population who should bear the expense and increased confusion?

On the other hand, if different projects do construct nonstandard time
scales to match their individual requirements - doesn't this really
indicate that selecting a standard time scale (whether UTC or TAI or
whatever) was NOT one of their users' requirements? Is the thought to
not only decouple UTC from synchronization with the Earth, but also
to somehow *require* the users of precise time to use only UTC (or
presumably approved derived time products)?

And to those of us on the outside of this debate, even the simple choice
of when to hold the debate is itself fascinating. Choosing to initiate
a discussion about this very long term concern in the few months between
the GPS week number rollover and the millennium Y2K rollover seems a
bit...overly confident.

The choice of GPS World as the forum for launching the discussion also
seems revealing. If we are to replace the spinning Earth as our master
civil clock, one would hope the change would be accomplished with all due
humility. Rummaging around the web for GPS resources (many excellent
sites) turned up this (similarly confident) comment:

    "There is always concern about the possibility of accidents
    whatever their cause, but it is hard to envision one caused by a
    GPS navigation receiver failing. No prudent navigator uses a single
    navigation aid and so no prudent navigator should be subject to an
    accident caused by a GPS failure. Certainly someone who misused any
    navigation aid might get in some sort of trouble should their sole
    means fail and under some circumstances that might happen with an
    old or new GPS receiver that failed for any reason. GPS is not
    allowed for use in private or public aircraft as a sole means of
    navigation, so even if a GPS receiver should fail there should be
    no serious problem. Hikers should always carry a map and compass,
    in addition to a supply of new batteries for their GPS unit."

    [Peter H. Dana, August 1999, "GPS Week Number Rollover Issues",


    "Secondary users, such as those who fly as passengers on aircraft
    equipped with GPS or who depend on telecommunications systems
    synchronized with GPS might want to asses [sic] the possibility of
    potential problems and make informed decisions, but should remember
    that no complex system offers guarantees against failures."

How might "secondary users" (that is - people) reasonably assess the
possibility of precision timing problems in a world that includes no
clocks tied to the rotation of the Earth? What informed decision is
possible when all clocks are referenced to complex systems subject
to failure? Upon what supplementary navigational aids will prudent
navigators rely?

Rob Seaman
National Optical Astronomy Observatory
Received on Fri Jul 21 2000 - 16:18:09 PDT

This archive was generated by hypermail 2.3.0 : Sat Sep 04 2010 - 09:44:54 PDT