DATE-OBS='31/12/99'

[Peter Bunclark]

What is happening about the millenium?  Is there any proposed way of dealing
with the 2-digit original spec. of the year of observation.

At the RGO, we have already digitized plates over a century old, and so the
problem is current.

One would not want (nor be allowed) to invalidate FITS readers expecting
the original DATE-OBS keyword, so a suggestion is
DATEOBS = '19960624'               / YYYYMMDD   Date with full year length.

This format will keep us going until 9999 Dec 31, by which time we will
have upgraded to FITS (Flexible Intergalactic Transport System).
It is big-endian, so can be sorted ascii-betically;
it is is a string so that a reader can go on to parse the year, month, day items; 
it is human-readable, and so complements Julian Date which is unambiguous.

Another possiblility would be to complement DATE-OBS with CENT-OBS = 20, ie
record the century of the observation.  Does 20 mean this century or next?

OBSERVER= 'Pete'              / Peter Bunclark.