leap second policy making

From: Rob Seaman <seaman_at_noao.edu>
Date: Fri, 14 Jul 2000 15:57:31 -0700 (MST)

Hi,

I have some concerns about the process being followed in the debate
about the leap second issue. My comments below tend to be framed in
an astronomical context, but clearly other interests abound.

Demetrios Matsakis says:

> The implementation guidelines followed by the IERS, and the authority
> of the IERS, came from the International Consultative Committee for
> Radiocommunications (CCIR). Therefore it is reasonable to assume that
> the International Telecommunications Union (ITU), which is the follow-on
> group to the CCIR, would be the body to decide if any change should be
> made. While not bound by the rules of other scientific organizations,
> they will be sensitive to the concerns of other international agencies,

The specific implementation of leap second scheduling presumably would
indeed depend on the deliberations of such mid-level international agencies.

This is very different, however, than stating that such agencies are
the ultimate authority for determining the fundamental international
policies constituting civil time.

National and international law governs fundamental calendrical decisions,
for instance. This may extend to local citizens as far as ballot referenda
for state holidays and such. It would be naive to state that government
and legal oversight play no role in establishing the policies governing
civil time. Any proposal to abolish leap seconds completely would
clearly embody a fundamental change to civil time. Merely changing
the scheduling algorithm for leap seconds might not.

Can anyone provide a historical perspective on the shift from ephemeris
time to universal time three decades ago? Presumably since this change
preserved the fundamental character of civil time to track the Earth's
rotation, no high level government actions were required?

> Report of the URSI Commission J Working Group on the Leap Second

A fascinating document. I presume the intent of the survey was merely
to begin to enumerate issues (both positive and negative) contingent on
any change to leap second policies.

> An e-mail survey to find possible adverse effects of a redefinition
> of UTC has identified some possibly expensive or unsolvable problems
> involving software rewriting or checking, which are listed below.

Is international policy really to be governed by the expense of
rewriting software? Isn't this the bullet we just dodged with Y2K?

The question should be - what is the best implementation of UTC? Only
then should the software and other costs be considered. Only under the
most extreme circumstances should a preferred implementation be hobbled
in order to mitigate technical issues - especially short term issues.

> Although it was not possible to quantify the financial scale of
> resolving the software problems, the largest expenses appear to be for
> satellite systems, of which one estimate of several hundred thousand
> dollars was supplied.

Ok - let's try to quantify the expense to astronomy. One might reasonably
expect that the cost of retrofitting astronomical software to handle a
non-earth centered UTC would mimic the cost of Y2K remediation to the
community. Astronomy employs a core group of a few hundred programmers
worldwide. One might estimate that - say - two hundred programmers spent
10% of their time working on Y2K issues over about 2 years. (Feel free
to supply your own numbers - these seem quite conservative to me.) That
makes 40 FTE-years. Even in the underpaid ranks of academia that's a
couple million bucks (_at_ $50K per FTE).

Astronomy is uniquely dependent on UTC, so the dollar amount may actually
be much larger. Virtually every telescope control system on the planet
is a unique software product. Telescopes tend to be pointed using fairly
simple minded assumptions about time, and UTC is often the underlying
system of time. (Prior to a few years ago, what else was easily available
on isolated mountain tops?)

On the other hand, some astronomical applications do require very precise
timing. These may be even more expensive to update to some new standard.

In a broader context there is also the issue of embedded systems. What
fraction of Y2K remediation costs involved embedded applications? Far
more such systems include clocks than include a knowledge of the date.
The issue isn't only whether these systems will require significant
changes to support any new civil time standard - the issue is also the
cost of simply vetting all those systems for potential problems.

> UTC (Coordinated Universal Time), which the public commonly confuses
> with Greenwich Mean Time,

No. The word "confuses" is very slanted here. There is no reason the
public *needs* to understand the difference. The "no leap second"
option (not the only proposal, of course) implicitly requires that
six billion people (and 100's of billions to follow) understand that
technical distinction in the future.

I would substitute "equates" for "confuses" with the understanding
that UTC ~= GMT is a useful approximation for many purposes.

On the other hand, I've seen several instances of professionals confusing
secular with cyclical effects during the previous leap second discussions
on various forums. Yes - the public manages to handle daylight savings
time adequately well twice a year, and yes - the equation of time does
vary by many minutes over the course of half a year.

But these are separate issues from allowing UTC to diverge monotonically
from UT1 (or roughly so) for the rest of eternity. (Yes, yes - I know
that it is UT1 that would be diverging from UTC.)

> A segment of the international timing community has proposed a revision
> of the definition of UTC to avoid the discontinuities due to intermittent
> leap seconds. A discussion of the motivations for a change and of
> possible solutions has been published by McCarthy and Klepczynski in
> the Innovations Section of the November, 1999 issue of GPS World.

Is this really the journal of record for the time and frequency community?
I'm fascinated that debate over such a fundamental international policy
change would be initiated through an op-ed piece in a minor publication.

I first heard about this initiative last November through a rather
tenuous string of colleagues. As far as I can tell, NOAO never received
an official survey mailing. It seems quite likely that the address
list used to distribute the survey pointed to a large number of public
information offices rather than to the scientific or technical staff
of various organizations.

Also, in the intervening months this is the first time that I've heard
about this article. (Thanks for publishing the PDF file, by the way.)
I would recommend distributing the article more widely at this point.
The survey was only the first step in educating the public.

> The authors consider the most significant reason for a change to be
> keeping spread-spectrum communication systems and satellite navigation
> systems compatible with each other and with civil times.

Worthy goals. But worth totally redefining the nature of civil time?

> The goal of the questionnaire was to find and categorize those operations
> that would be adversely affected should a change in UTC's computation
> be made. The questionnaire focused on the possibility of simply inserting
> no new leap seconds, although alternative solutions were also solicited.

Why focus on only one option out of five?

> The principal object of the questionnaire was to find what systems
> would be adversely affected should a change be made in leap second
> procedures rather than to convince users of the need for a change or
> to take a vote.

Excellent goal. Are "systems" and "users" the right metaphors for a
policy change that would effect every human usable clock on the planet?

Abolishing leap seconds is not principally a technical issue. Modifying
civil time for everybody is not a decision only for experts. (But changing
the scheduling algorithm might indeed require only purely technical oversight.)

> we caution that this was by no means an unbiased sampling of all who
> would be affected by the change. All responses were counted only once,
> with preference to the most practical grounds for objection.

It seems to me that the principal goal of the survey was to list areas
requiring further study. If a single "user" suggested two problem areas,
did both make it onto the list?

And what were the actual grounds for objection? It would be quite
worthwhile for such case studies to be discussed in detail on this forum.

> Twenty-six others indicated that software would be a serious problem - a
> very few of these were from people who did not seem to understand the
> proposal.

Is there an actual proposal on the table? You might get more informed
responses if a straw proposal (e.g. - "After 1 January 2005, no further
leap seconds will be issued. UTC will be allowed to diverge from UT1.
The GPS offset will be adjusted to match UTC on 1 January 2006." - or
whatever) were circulated through channels more official than GPS World.

> There were 9 responses involving telescope control; one of these,
> from the Keck Observatory, provided a rough estimate of a few
> programmer-months.

Note that some legacy telescopes might well be closed rather than invest
the expense of such a TCS rewrite. On the other hand, time at large
telescopes is valued at thousands of dollars per night - and any control
system rewrite may require several engineering nights for verification
and testing.

Contingent costs should be included in any budget impact analyses.

> Others pointed out the problems computing eclipses and occultations,
> for telescope pointing by amateurs, or with code they had themselves
> written for professional-level projects such as speckle interferometry.
> One observatory indicated its station clock can not accommodate a large
> UT1-UTC correction.

I'm sure this represents only a tiny fraction of astronomical issues
that would be involved with abolishing leap seconds. Note that one
second of time is 15 seconds of arc at the equator. Only a decade or
two of divergence would represent greater than the field of view of
many current instruments. Even the most rudimentary telescopes and
inexperienced amateur observers will easily detect the clock divergence
after just a few years.

> Fourteen more indicated that software issues would be a problem, but
> that they are probably solvable.

Solvable, but expensive. And the expense would also have to be added
to future projects - no future astronomical software could assume a
naive model of time.

> Ten responses involved navigational software. Taking the example of a
> software product of the US Naval Observatory (USNO), pilots and sailors
> are given the option to input UT1-UTC. However, it is expected that
> many users would not understand this and enter 0, leading to noticeable
> errors within a decade. These are similar to the telescope-control
> problems covered above, except that one could not and should not expect
> the general public to ever understand these issues.

Realistically one could not even expect professional astronomers and
other scientists to ever completely handle these issues properly.

> Not tabulated is an informal comment seriously made to the Working
> Group's Chairman, by a respected and competent scientist from a
> non-western nation, that astrologers would be adversely affected.

Why not tabulate this? Many aspects of the proposal (well, "suggestion")
appear to be governed by economic issues. One might well expect that the
total economic impact of astrology is greater than that of astronomy...

> Although one of these responders feared governments would not follow
> the scientist's lead, we find it difficult to believe that governments
> would, on their own, choose to add leap seconds.

Which scientists? Of course the governments would not manage leap
seconds directly - but they certainly have the authority (and funding)
to implement such civil policies indirectly.

> Three thought we should not adopt a system which will fail in the long
> run, even if that is a very distant time in the future. (It could be
> pointed out that all current time systems will eventually fail. Well
> before 2050 we could be routinely adding more than one leap second per
> year, and when we reach the point where a day is 48 hours long we
> would have to add a leap second every second. Even the Gregorian
> calendar will eventually need revision because in a few million years
> the Earth will rotate less than 365 times per year, and leap days will
> not be necessary.)

There is a huge difference between 2050 and "a few million years".
21st century policies should be based on 21st century trends. All
solutions are interim. The most important constraint may be to implement
the next interim solution in a way that will support future changes.

> Thirty-eight expressed opposition, but gave no specific reason.

And a good follow-on survey would be to ask them why.

> Some of these pointed out that TAI was readily available, or indicated
> that they had seen no justification for a change. (We had intentionally
> provided no justification in the initial questionnaire, but two responders
> replied that they still believed there was not enough reason to change
> even after having read "standard response".)

And perhaps this also says something about the standard response? The
standard response mentions the GPS World article in passing. Perhaps there
were also other discussions prior to this article? The full context of
this initiative should be made clear.

Was the intent of providing no justification to avoid biasing the survey?
A better procedure would be to make the best case possible for various
scenarios and then allow both opponents and supporters the chance to make
or break each option.

> One suggested that a redefined UTC should have a new name.

This has historical precedents and would minimize confusion and permit
easier explanation to "civilians" should any major changes be made.

Personally I vote for leaving UTC alone except perhaps for tweaking
the leap second scheduling algorithm.

> Forty-eight responses in favor were received, several from people who
> experienced minor problems now in handling leap seconds, such as
> confusion due to GPS time being currently 13 seconds offset from UTC
> and computer errors at the time of the leap seconds.

Perhaps such favorable responses say more about GPS than UTC.

> Along with responses based upon reasons already covered in the GPS
> World Article, there were also three from the highly undersampled
> group of computer programmers,

"Highly undersampled group" is a value judgement, not a statement with
any statistical merit. This is not a survey of personal preference and
it does not fundamentally matter how many folks "vote" for any particular
result (now that my own vote has been registered :-)

What is the best design for UTC and civil time? I suspect that the best
design does not depend on current technical headaches. GPS and NTP and
WWV are interim solutions themselves that are meant to bring civil time
to civilians. The needs of the civilians come first. Professionals can
always adapt to any self-consistent standard. If the GLONASS and spread
spectrum communities are encountering some difficulties adhering to the
current standard - perhaps they are the ones who should bear the expense
of resolving their problems?

What standard for civil time best supports the civilian population?

Rob Seaman
National Optical Astronomy Observatory
Received on Fri Jul 14 2000 - 15:57:39 PDT