A warning for FITS y2k implementors

[William Thompson]

Don Wells <dwells at NRAO.EDU> writes:

>An implementor of FITS y2k code told me:

>  "..there was one sort of amusing problem with the initial
>  implementation that other systems might run into. By default with the
>  new-style we include the time. But because of floating point rounding,
>  the new style date would turn into previous_day:T23:59:59.99 or
>  whatever. While this is fine for systems that read the time part of
>  the new format, it would presumably cause systems that ignore the time
>  part to think the date was a day earlier than it actually was.."

Something else that implementers of the new standard should be aware of is that
under certain circumstances, a time such as 

	1997-06-30T23:59:60.999

could also be a valid time string, on dates containing leap seconds.

William Thompson