SED data model v0.92

[David Berry]

Doug,

> > Am I correcting in thinking that if the spectral coordinate
> > is linear with channel number, you still have to supply all the spectral
> > coordinate values explicitly, rather than just giving a scale and offset?
>
> Yes, this is the case at present.   The main reason we did this was
> to avoid getting into all the issues involved in specifying a general
> spectral WCS.

Would you really need to do this up front? The approach Jonathan I have
been looking at in the IVOA Mapping class (see
http://hea-www.harvard.edu/~jcm/vo/docs/map.pdf)
is to view a Mapping as a black box which has a certain number of inputs
and a certain number of outputs (one of each in the case of simple
spectra). The recipe which is used to transform an input value into an
output value is encapsulated entirely within the Mapping object. Using
this idea within the spectral data model would just mean providing a place
holder for a Mapping object somewhere. The only thing which you need
specify about the Mapping in the spectral data model schema is that it has
one input (corresponding to channel number) and one output (corresponding
to the spectral coordinate value). Is there any need to specify anything
at all about *how* the output is derived from the input (i.e. about the
nature of the recipe encapsulated within the Mapping)? Can that not be
left up to the person creating the spectrum? Do we need to follow the
FITS-WCS route of demanding that only a small number of "blessed" recipes
be allowed?


David