[fitsbits] WCSAXES usage

[Doug Tody]

Hi Bill -

For backwards compatibility I believe we had to define that WCSAXES
is not required when WCSAXES=NAXIS.  Evidently the standard allows
the keyword to be omitted when WCSAXES > NAXIS as well.  However I
suggest that it "should" be provided in this case (although FITS
does not make such a distinction).  It might be nice to issue an
informational warning in such a case, even though it is legal usage.
That is, flag any actual errors, but also issue informational comments
when there is a departure from suggested usage.

 	- Doug


On Thu, 5 Aug 2010, William Pence wrote:

> I'm in the process of updating the FITS verifier code to deal with
> changes in version 3 of the FITS Standard, and have a question about the
> correct usage of the WCSAXES keyword: is this keyword optional in cases
> where the WCS has more axes than specified by the NAXIS keyword, as in
> the following fragment of a FITS image header:
>
> NAXIS  = 2
> NAXIS1 = 512
> NAXIS2 = 512
> CRPIX1 = 256
> CRPIX2 = 256
> CRPIX3 = 1
> ...
>
> Should the FITS verifier issue a warning in this case (where there is no
> WCSAXES keyword) that the index on the CRPIX3 keyword is greater than
> the NAXIS keyword value?
>
> The FITS Standard states that the default value for the WCSAXES keyword,
> if it is not physically present, is "NAXIS, or largest of WCS indexes i
> or j".  This suggests that the above example header is legal, and that
> the default value of WCSAXES is 3 this case.  If this is the correct
> interpretation, then the FITS verifier should not issue a warning in
> this case.
>
> Bill
>