[fitsbits] 64-bit integer comments

[Thierry Forveille]

Mark Calabretta writes:
 > While we're counting bits, I note that the precision of keyvalue data
 > types is essentially undefined.  Specifically, I believe that most
 > generic header parsers that encounter
 > 
 > PSRTIME =       88402115102000 / [us] Time since T0
 > PULSENO =           9465023873 / Pulsar pulse number since T0
 > 
 > will pack them into 32-bit ints and lose essential precision.  The FITS
 > standard is silent on this issue, perhaps it should address it also. 
 > 
Not sure that I like this idea. Currently FITS is a specification for
data writers, and has no notion whatsoever of a "conformant FITS reader".
That's fairly natural for an interchange format where implementation of
the full standard is not required, and changing that feels like a slippery
slope. If nothing else, it would considerably raise the bar of FITS support
costs for smaller package. Essentially, we'd be telling people who might
never need to read a PSRTIME that they must be able to do so.