Re: "A diagram of C23 basic types"

On 16/04/2025 02:53, James Kuyper wrote:

On 4/15/25 18:56, Keith Thompson wrote:
...

The uncertainty in the timing of January 1, 1970, where 1970 is a
year number in the current almost universally accepted Gregorian
calendar, is essentially zero.

 Modern Cesium clock are accurate to about 1 ns/day.That's an effect
large enough that we can measure it, but cannot correct for it. We know
that the clocks disagree with each other, but the closest we can do to
correcting for that instability is to average over 450 different clock;
the average is 10 times more stable than the individual clocks.
 Note: the precision of cesium clocks has improved log-linearly since the
1950s. They're 6 orders of magnitude better in 2008 than they were in
1950. Who knows how much longer that will continue to be true?
 

I don't think cesium is still the current standard for the highest precision atomic clocks.  But anyway, the newest breakthrough is thorium nuclear clocks, which IIRC are 5 orders of magnitude more stable than cesium clocks.  (And probably 5 orders of magnitude more expensive...)

...  Same for any other less commonly
used chosen epoch.  The fact that the number 1970 is arbitrary
is not a problem for software.  In fact it's an advantage, since
there's no uncertainty in the presence of any new information.

 I agree, which is why I identified that epoch as the one I preferred
over both of those.