Subject: Re: Representing Time
From: rpw3@rigden.engr.sgi.com (Rob Warnock)
Date: 1999/03/19
Newsgroups: comp.lang.lisp
Message-ID: <7csevv$eue5o@fido.engr.sgi.com>
Kent M Pitman  <pitman@world.std.com> wrote:
+---------------
| ...a conforming Lisp doesn't come with the necessary hardware to notice
| the slowing of the earth and to guess that probably a leap second was added.
+---------------

Is that a defect of the physical Earth, or of the politics which defined
a "conforming Lisp"?  ;-}  [Sorry, couldn't resist!]

+---------------
| (Also, it *is* political when to add the second, since the slowing is
| something that is happening continuously, not at the point of time the
| second is added.)
+---------------

Actually, not quite. If the Earth's rotation were *predictable*, they could
set up a formula similar to the leap year formula that said "add leap seconds
*here*", and we could all write it into our time library codes (and our
definitions of software "conformance") the way we do the leap year adjustments.

But in fact (mainly due to somewhat chaotic variations in the circulation
of the molten iron core deep inside), the Earth's rotation *doesn't* seem
to be very predictable, at least, not at the fine-grain needed for choosing
leap seconds.  So what they do is *measure* the rate, and when the discrepancy
between the rotation and "atomic time" gets "large enough" (by some rules
that were written down when Coordinated Universal Time was defined by the
ITU in 1970), they schedule a leap second at the next standard opportunity
(June or December) so that the difference bewteen UTC (cesium beam atomic
clock mean time) and UT1 (Earth rotation time) never exceeds 9/10'ths of
a second.

So, yes, again, there was politics involved when Coordinated Universal Time
was set up in 1970, but since then they've just "turned the crank" on the
rules for adding leap seconds. By which I mean, no *new* politics (which
is what your statement implied to me) is currently involved in deciding
when to add a leap second.

+---------------
| Daylight time is not a problem, by contrast, because machines largely
| can be programmed to anticipate the occurrence of daylight time.
+---------------

And if your machine *did* have equipment for measuring the Earth's rotation
(say, an automatic star-tracker) and if it had the "rules for adding a leap
second" [set in 1970] programmed into it, it could anticipate the need for
a leap second as well as the ITU and the "UTC" consortium... but unfortunately
neither can anticipate the future need for a leap second very well! This
old Earth just ain't very cooperative!


-Rob

p.s. Imagine if the variation in the Earth's rotation were several orders
of magnitude higher, e.g., so that the number of days in a year were a
random variable with a mean of 365.24 but a standard deviation of, say,
3 or 4 *days* (instead of the current SD of less than a second). How might
*that* affect our notion of computing with "time" (not to mention social
issues)?  [Hmmm... Sounds like an idea for a science-fiction novel in there
somewhere...]

If abyone can think of a good solution for *that* problem, it would probably
work even better for our actual current situation.

-----
Rob Warnock, 8L-855		rpw3@sgi.com
Applied Networking		http://reality.sgi.com/rpw3/
Silicon Graphics, Inc.		Phone: 650-933-1673
2011 N. Shoreline Blvd.		FAX: 650-964-0811
Mountain View, CA  94043	PP-ASEL-IA