[fitswcs] NAXIS vs. WCSDIM

[Doug Tody]

On Wed, 26 Sep 2001, Arnold Rots wrote:
> I remain unconvinced by Doug's arguments.  If I understand it
> correctly, acceptance of the WCSDIM keyword will not change anything
> without substantial software changes: those who currently use
> degenerate axes will be able to continue to use them but not
> understand the WCSDIM convention, and those who do not currently use
> degenerate axes will adopt WCSDIM but still not be able to read
> interpret images which use degenerate axes.

Arnold, this is not quite correct.

Those who write images with degenerate axes will be able to continue to do
so, but they will be able to read any image with a valid WCS since WCSDIM (or
WCSAXES or whatever we call it) would be part of the new WCS standard.  We
are still defining the WCS standard and we will all have to modify our WCS
software, to some extent, to support whatever is finally adopted.  Using
WCSDIM rather than NAXIS to determine the WCS dimensionality should be a
pretty minor change to existing software since the WCS formulation itself
changes very little.

Current software which does not support images with degenerate axes would
continue to have this problem regardless of what we do with WCS, since this
an issue with the image pixel matrix rather than the WCS.  However, this is
already the case today, and indeed has been for years, and is in general NOT
a problem.  Only a fraction of the software out there needs to be able to do
general multiwavelength analysis.  If we start routinely generating "higher
dimensional" images however, potentially any program, including old ones that
have been working fine for years, might be forced to deal with such data.

The issue isn't just minimizing changes to existing software however.  Some
of us feel that having the image and the WCS stand alone as separate objects
(while still being easily associated) is the more general and flexible
approach in any case.

	- Doug