[fitsbits] 64-bit integers

[Thierry Forveille]

William Pence writes:
 > The question I would like to raise is, should a proposal to support 64-bit
 > integers in FITS now be drafted and submitted to the 3 regional FITS
 > committees for their consideration?  Are there any compelling reasons not to
 > do this now?  
 > 
Could you come up with a few practical examples that actually demand the
dynamics of 64-bit integers? 

My view of the issue is that convenience to writers is not a sufficient 
argument to add a new datatype to FITS, given the cost to readers. The 
reason has to be some dataset that cannot be coerced into existing datypes 
without loss of information. I cannot myself think of actual examples that
don't fit inside the dynamics of 64-bit floats (which can be used as
~~48-bit integers; I don't remember the actual number of mantissa 
bits...), but that may just reflect my personal biases (radio+IR+optical,
but no high energy).

Best regards,
Thierry