[ntp:questions] A Suggestion For Abolishing the Leap Second

Quadibloc jsavard at ecn.ab.ca
Sun Jun 3 11:53:16 UTC 2007

Brian Garrett wrote:
> http://iraf.noao.edu/~seaman/leap/ contains one of the better discussions of
> this issue I've seen.

This one also contains a proposal to change how time is kept: insert
leap seconds on a *monthly* basis so as to keep UTC within, say, 0.6
seconds of mean solar time.

> http://www.cl.cam.ac.uk/~mgk25/time/utc-sls/ outlines a similar proposal to
> Mr. Savard's.

The proposal is similar to mine in one way, but it's actually quite
different. He proposes that if a leap second is to be added to
December 31st, for example, then one millisecond should be added to
each of the last 1000 seconds of December 31st.

This would mean that there would be no 11:59:60 second in time
signals. Otherwise, however, it doesn't eliminate the problems
associated with leap seconds. So, if you forget to manually tell your
local clock that there's a leap second this year, *even* if it's just
a plain old quartz clock, on December 30th it's keeping perfect time,
and on January 1st it's a second out... just as is the case under the
current system.

If I instead take those 1000 milliseconds, and give one of them to the
last second of every eight hours, spread out through the first 333 1/3
days of the year, as I now propose, then periodic checking of a quartz
clock against the time signal, as often as is necessary to keep its
time within whatever tolerance is required by its application will
*also* take care of the leap second to the same tolerance.

And I've even come up with a refined proposal. After all, if it *is*
necessary to change UTC, this means a can of worms is opened up. So
the people who were left unhappy the last time it was changed will
perhaps ask, why can't we go back to making us as happy as we were
before the current system was adopted? So, after thinking about
things, I've come up with an even *more* improved proposal that makes
the people navigating by sextant and compass as happy as they were
before - no more DUT, no more UT1 - UTC correction to bother with.

Lets have steps that are multiples of 100ms applied on a monthly basis
- and spread them out in the form of adding one millisecond to the
last second of every six hours for the first 25 days of the month.

The Internet is happy. Computers are happy. Seconds just come in
regular succession with no leap seconds, and periodically setting your
(low-accuracy quartz) clock by WWV invisibly takes care of the
discrepancies that used to be acknowledged in the form of leap

For stretches of 5 hours, 59 minutes, and 59 seconds at a time, the
time signal is a constant stream of SI seconds, so people calibrating
electronic equipment are happy. (Of course, to know the span of 6
hours to which this applies, they need to remember the time zone
they're in.)

Astronomers and celestial navigators are happy.

I know that there are a lot of conflicting requirements involved in
all this, but it's still not quite a zero sum game. It is possible to
navigate through the set of requirements that is being advanced, and
find something that comes closer to what the different user
communities require, I think.

Of course, it's possible that a novel proposal such as mine might turn
out to cause problems for a user community nobody had heard from

John Savard

More information about the questions mailing list