Re: [LEAPSECS] the legacy of ephemeris time

From: Poul-Henning Kamp <>
Date: Mon, 22 Dec 2003 20:19:14 +0100

In message <>, Markus Kuhn writes:

>But what do you do when the offset between New York City local time and
>Univ^H^H^H^HInternational Time becomes significantly larger than 24
>hours, such that the discrepancy in date between TI and local time
>becomes a matter of days rather than hours, which would happen just a
>view millenia down the road?

A millenia ago, some of the grandparents of the people who fought
a Hastings had just started having kids.

Two and a half centuries ago, two weeks were taken out of calendars
here and there, in order to get things into better shape. Apart
from a bunch of nerdy brits, nobody was hurt in the process.

One century ago, Bohr and Einstein had still not done their bits.

Half a century ago, atomic timekeeping was very experimental.

Leap seconds are only 3 decades old.

Time and effort, such as the above, spent on pondering the inconvenience
potentially caused to civilizations a millenia, or even just a
century in the future, is at best wasted, at worst in the way of
our progress and the beneficial scratching of the current itch.

If we look back in history, there are a number of lessons to be
learned about timescales:

The lesson learned from 1752 (and forward) is that a large ad-hoc
change can be implemented easily, because it is so disruptive that
there is no choice but to pay attention.

The lesson from the "rubber-UTC" implementation before 1972 was
that we very much want seconds to be the same size.

The lesson learned from march 1st 2000 was that as long as the rules
are simple, predictable *and trigger often enough* people will get
it right. (The 400 year leap-year rule took surprisingly few people
by surprise.)

The lesson learned from leap-seconds is that infrequent, ad-hoc,
minor adjustments are a bad thing. Mostly they are bad because
they are not significant enough to be implemented in more than the
most demanding applications, and in these applications we curse the
fact that they cannot be predicted more than 6 months down the road.

Leap-hours would be too big to be ignorable, they will happen
infrequently enough to give the mass-media two free weeks of column
stuffing and they will not be inordinately troublesome as we already
know from from daylight savings time.

In fact, I'll be the first person to point out the advantage of
leap-hours: it would give the dairy farmers one less opportunity
to complain about the cattle not understanding time adjustments.

Alternatively, a century from now, the settlers on Mars and the
colonists on the interstellar ships will be most amused when they
read in their historical records that the reason why they have to
stick a second into their timescale tonight, is that a hundread
years ago, scientists were worried about the ability of New Yorkers
to figure out what date it was a millenium in the future.

When pondering the future of leap-seconds, the most important advice
for both camps was given by Robert Storm-Petersen many years ago:

"It is difficult to predict, in particular about the future."

Merry Xmas,


Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk_at_FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Received on Mon Dec 22 2003 - 11:19:29 PST

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