time zones, offtopic (Was: Various fixes for Celeste)

Stan Heckman stan at stanheckman.com
Tue Dec 7 06:06:32 UTC 1999


"Eric Ulevik" <eau at fast.fujitsu.com.au> writes:
> See ftp://maia.usno.navy.mil/ser7/tai-utc.dat for a table assigning
> all leap seconds.

This table carries us back to 1961, but notice that the offsets before
1972 are not discrete leap seconds, they are changes in the length of
the second. That is, these "leap seconds" are not assigned to
particular days, they are spread uniformly over time intervals of
months.

> These are convertible. Easiest way is to use Terrestial Time.

My understanding is that terrestrial time is clock time as measured on
the (idealized) surface of the rotating earth. This gets us modern
ephemeris times, but cannot be converted to older Julian day numbers,
because no one measured the length of those older days.

We have estimates of the length of historical days, but nothing as
officially blessed as the table of leap seconds. And we can expect
astronomers, historians, and marine paleontologists to continue
changing the estimates. So clock time is convertible into solar time,
but only by referencing the particular conversion we are using.  This
is definitely too complicated for me.

> Not really... leap seconds are fiddly. But I think they could be
> ignored in an initial implementation.
>
> I think ignoring historical dates would be a pity. An initial
> implementation could just support the propleptic Gregorian
> caldendar. Other calendars could be added later.

Okay. If we are going to choose historical dates instead of leap
seconds, perhaps method comments should indicate that historical times
measure position of the sun, while the modern times measure the rate
at which clocks run. We have two classes, Date and Time. Date can
unambiguously be the (possibly proleptic) Gregorian date. Time is the
difficult case. Perhaps Time's comment could read "...UTC after 1972,
solar time at Greenwich before." This should make it obvious that we
aren't supporting one second accuracy for durations extending before
1972.


-- 
Stan





More information about the Squeak-dev mailing list