HI Simon:
I am not sure I agree - in that one could determine the nature of the domain by
inspection of the schema in any event - however I agree that spatial-temporal
coverages have some operators that might not apply in general - not sure I can
think of any for spatial - but I am sure there are some that at least would be
more abstract . We can do netCDF over spatial-temporal domains now and I think
in the context of GML in JPEG 2000 this is a valid first step. Hopefully the
generalizations done to handle more abstract (i.e. non-spatial) domains can
then done as extensions of what exists in GML now.
Respecting observations - it is about the collection of things that could be in
themselves coverages (e.g. take a photograph) or as you note construct a
coverage on aggregation - of course not all coverages arise from observations -
they could come just from simulation or an analytical model.
Cheers
Ron
Sent: May 13, 2005 6:46 PM
Yes again - Coverage is clearly a restriction of a more generalised parent, perhaps
starting at "Function".
However, the restriction to spatio-temporal domains is useful.
This is a special, frequently-occuring subtype of the general concept, that is
important enough in GIS to warrant its own name.
There are certain operations that apply to coverages that do not apply to the
parent classes, and it is useful to have things classified in this way to allow
these operations to be available.
So, I think we do need a more general class, but we should also retain
"coverage".
netCDF is related to the parent of coverage, and JP2K relates to a restriction
of coverage.
And the Observation model relates to metadata about the estimate of values in
the range.
Simon
