more on FITS and VO
jcm at head.cfa.harvard.edu
Fri Nov 28 09:00:21 PST 2003
The example I find compelling is the case where an archive,
or cutout service, may return a null image because the data
either do not exist or have not been successfully processed.
In VO-like applications we may want to generate a lot of
data requests, some of which may return null. Sometimes it's
easiest to handle that by having your analysis script handle
empty images anyway (rather than having the VO return
error conditions that your script has to handle, making it
more complicated). As you point out, the null image's header
may contain useful information. Having the null image
on my disk, instead of having nothing at all, reminds me that
I didn't forget to query for it, I asked for it and got a
For services to have to return a different mime type in this
case seems annoying to me. I think the NAXISn=0 case should
be considered valid (if degenerate) image/fits.
Of course I'm a mathematician, so the concept of an image of
size zero holds no horrors for me :-). I don't think it
will break any applications though.
More information about the interop