datetime

[Rob Seaman]

Patricio Ortiz says:

> a "datetime" (however it is represented) corresponds to an instant in 
> time.

Well, simultaneity is not just an issue for GR.  A datetime, whatever 
time scale is used, only makes sense with respect to the location of 
the observation (or a derived location such as the barycenter).  Some 
representation with a complexity similar to STC is required to 
unambiguously capture that instant in full.

> Julian Date and Modified Julian Date (which do assume the usage of UTC)
> seem to be one of the most appropriate way of representing such instant
> in time. Using the ISO standard (string representation) is quite 
> common,
> but any application wanting to compare instants in time needs to 
> convert
> it to a floatinng number (double).

If the precision required is larger than a second (a typical case for 
ground based observations), an integer (long) may be preferable.  A 
simple epsilon test may be sufficient to test for equality.  A more 
general case may be constructing a histogram representing the passage 
of a wavefront (or non-EM event) - in that case we may want to consider 
more obscure questions like whether the histogram intervals are 
half-open on the top or the bottom.  Deciding the simultaneity of two 
events that are extended in time rather than instantaneous is a more 
interesting question than simply whether two intervals overlap.  (Even 
ignoring the differing physics of signal generation and propagation 
that may apply to two different wavelength regimes.)

Rob Seaman
NOAO