This is not an ideal solution, and I welcome comments on how we can
go about changing this. Rectifying the grid to evenly spaced means a
change to the original dataset (though it is necessary to make
geotiff).
Hopefully this will be something that we can push into the GML in
JPEG2000 coverage format.
Norman
I feel uncomfortable about a WCS that more or less silently does an
interpolation. It also contradicts the WCS preamble that stresses the
original data quality. IMHO any interpolation should be explicit to
the requestor. So the consequence for me is that functionality should
be available that allows to generate inbetweenings.
My suggestion is to put such functionality not into a WCS 1.1, but
into the WCPS (Web Coverage Processing Service) recently proposed
where a coverage language allows to arbitrarily combine request
facets. WCS I think would suffer from too much changing it (some
reasons: we see change requests coming up in line, with many desirable
operations - where is the limit, are they mutually axclusive or
combined, and in particular: how stable can WCS get?).
-Peter
I agree strongly that WCS should be capable of serving out the data
unchanged. This is what attracted Unidata to trying WCS. The WCPS
proposal seems on the surface a reasonable way to factor things, but i
havent had time to understand it in any detail.
After working with WCS a bit, I was getting the feeling that the
designers were greatly influenced by the WMS, which allows clients to
specify exactly what they want (resolution, bounding box, projection,
etc). This is a nice feature for clients, for sure, but not so easy for
servers, and not focused on serving out the raw data. And of course as
you point out, there are so many services a client might want.
In pursuit of that, for example, our server can send out geotiff with
the data in floating point "Data Sample" format, even though geotiff
readers are not required to be able to read this.
Anyway, Im glad that others are also interested in focussing on
"original data quality".
