UTYPE: Comments on Alberto's note

[Jonathan McDowell]

Alberto's UTYPES
----------------

What I think your UTYPES provide is a way to refer externally to an
attribute in a data model. In a pure XML serialization, internal
cross-referencing would be done in a standard way using e.g. XPATH. Or
to say it another way, the XML-specific UTYPEs are just the XPATH name
and are fixed by the names we give to the tags in serialization. So the
way to make this synch up is to require the external UTYPES and the XML
tag names to be equal. It doesn't matter which we specify first. In that
context, what you are proposing is XML element and attribute names for
the various pieces of the coverage model.

What you are doing in your document is remodelling the CoordArea
concept from Arnold's STC document. Compared
to his document, you add the following (which I like):
  - regions defined by a name via a lookup (resolver)
  - representing an interval by [start length] as an alternative to [start stop]
and you do not allow 2D regions (circles, ellipses, annuli) except
for polygons. I do think we need more than polygons.

Some specific comments:

I would sort things the other way around:
  coverage.position.spectral
not
  coverage.spectral.position

because I think by default in this N-dimensional space,
within which 'position' is one concept and the spatial,
spectral parts are just projections.

I also don't think 'position' is good English for the concept you are
discussing, for me 'position' does not include the idea of extent or
diameter, and doesn't work well for the spectral case; 'region' is a
better match (and I have argued for 'support' as a precise mathematical
term, but I could live with 'region').

 - Jonathan