[LEAPSECS] Lets get REAL about time.

[Zefram]

Poul-Henning Kamp wrote:

>The Epoch is not important, and for reasons of debugging programmer

>mistakes, preferably clearly visibly different from other magic numbers.


I think he was suggesting that you use the 1958-01-01 epoch that NASA (and
I, in my desktop clock program) already use.  That would make TAI times
clearly distinct from Unix times, while providing a slight convenience
among people who already count TAI seconds linearly.  Your proposed epoch
achieves the former but not the latter.  There's *also*, if you're going
to use floating point, a slight advantage in the epoch being relatively
far in the past, so that you don't go through a great many changes of
exponent in the near future.

-zefram