[fitsbits] 64-bit integers

[Thierry Forveille]

	Hi Greg,

Greg Hennessy writes:
 > In article <15261.21485.339424.184013 at hoku.cfht.hawaii.edu>,
 > Thierry Forveille  <forveill at cfht.hawaii.edu> wrote:
 > > Could you come up with a few practical examples that actually demand the
 > > dynamics of 64-bit integers? 
 > 
 > The FAME astrometric mission will have a 10Mhz clock operating for 5
 > years. That will be about 2^54 clock ticks, which can't be stored
 > in a double precision without loss of precision.
 > 
Time indeed is a quantity that taxes dynamic range limits. The traditional
answer has been to cut it into pieces (such Julian Day plus Time of day,
or integer seconds and fractional seconds, or whatever similar combination)
that individually fit the range restrictions. I can see some merit in the 
simplification of dealing with a single quantity, to be balanced against
the added complexity in every fit reader. If that's the only case I'd
probably still tend to vote to keep 64-bits integers out for the time 
being.

Thierry

PS: Please transmit my salutations to all the NPOI/FAME folks, specially 
at the present difficult time. When you see them that is, as presumably 
nobody has been allowed in at USNO today.