On 7/25/2011 11:44 AM, Jim Biard wrote:
Given all the various complications described, perhaps it would make
sense to make the "band" coordinate variable an array of label
strings. The label strings can differentiate the bands without
imposing any sequential numeric sequence constraint.
A string-valued coordinate (aka a nominal coordinate) would be ok, but
it seems like the common case is that there is a numeric coordinate that
could be used, which i think would be preferred. I wouldnt let the
unusual cases dominate the convention.
An "auxiliary coordinate variable" doesnt have any numeric constraints.
Other variables would be used to specify wavelength, wavenumber,
bandwidth, polarization, etc, and we could define an attribute to
attach to the band coordinate variable that would identify the group
of variables that are being used to describe the bands. (This would
be analogous to the coordinates attribute, but would be applied to the
coordinate variable instead of the data variable.)
To use existing CF convention, youd just put them into the coordinate
attribute. You could propose a new convention.
