[fitsbits] Re: Reading floating point FITS files

[Thierry Forveille]

Mark Calabretta writes:
 > >Sure, but how many users are there that care about the data 
 > >values/detailed coordinate information and that would not want 
 > >to play with their own colour tables?
 > 
 > Currently, there is not even a way to indicate in FITS that a particular
 > image (e.g. the crab nebula multispectral composite) is meant to have a
 > colour interpretation.
 > 
Sure, but does that need to be supportable in FITS at all? 

 > > My guess is that there are large constituencies for each
 > >capability but that they are mostly disjoint. They might even be the same
 > >person by the way, just not at the same time: I am glad to use nicely
 > >prepackaged pictures for my classes, but for my research where I use the
 > >full data I'd play with the colours anyways to look for additional 
 > >structure.
 > 
 > Does this mean that you store the image data as jpeg, FITS, or both?
 > 
In one case jpeg (or Postcript) is fine, in the other FITS is fine. For me
the two cases are essentially non-overlapping, but when they are not one
can use both formats side by side on the same web site.


 > Of course, you would still be able to twiddle the colours in a colour
 > FITS dataset; 
I do realise that the mechanism you propose would only provide a default
colour rendering (or perhaps several) for the data set.

 > > Most astronomical datasets have more information than can be
 > >displayed on a screen no matter how smart the lookup table choice, so
 > >we'd just make one particular reasonable choice. That's fine if it
 > >can be done (very) simply, but otherwise I'd rather stick to jpeg (or
 > >postscript, or whatever) stored next to the fits file on the web site.
 > 
 > You would almost always want to do that anyway because the general public
 > cannot be expected to have ready access to FITS viewers.  However, that
 > is not the point.
 > 
My point is that people who do have access to FITS viewers probably
do not care too much about being provided a default colour rendering.
Of course I am just one single data point and perfectly willing to be 
outvoted, but for now I haven't seen an outcry of users begging for
the feature. Also, my position would depend on the complexity of the
implementation: if the cost is small, the usefulness standard can
be lowered a bit...