This is a multi-part message in MIME format.
While there are many different understandings/view points/etc of feature and coverage, I would like to point out that
the main reason for "coverage" service is to provide "GRID coverage" whose data values are
"regularly" aligned with some axes/dimensions so that one can specify low/up limits and an interval along
each axis/dimension and get an array of (many) coverage data points. By "regularly" I mean that the coverage
data points have the same interval along each axis/dimension. If an axis is a well defined coordinate axis, such as
geographic longitude, then the data points are said to be rectified along this axis. If an axis is something like a
satellite track, the data points may still be considered rectified along this axis, but since a satellite track usually
does not align with any geographic/projected coordinate axis, the data points are thus usually considered un-rectified
(with regard to well defined earth coordinate reference systems). Whether rectified or not,
a grid coverage data is a n-D data array with all data points having the same data type.
In case a coverage contains multiple "range sets" such as temperature,
humidity, and land surface type, a coverage will have multiple arrays each having the
same data type. This well maps to the variable array(s) in a netCDF file.
This spatial (or geometry) property of coverage is fundamentally different from most feature types. I consider a road feature a typical feature type. To describe a road feature, we usually use discrete pairs of coordinates (x,y) to indicate its spatial location and use attribute values such as pavement type, route number, and road width. The attributes may or may not have the same data type and a feature usually has relatively fewer values for each attribute as compared to data point values in a grid coverage's range set. The data structure of a feature is thus usually different from a grid coverage.
Due to such major differences, it is inevitable that WCS interface will be
different from WFS interface.
There are, of course, cases where feature and grid coverage are more similar, such
as John Blower's point time series profile case. When considered as a grid
coverage, it has only 1-dimension and the intervals of data points along this
dimension may often be irregular (e.g., 50mb, 100mb, 200mb). When considered as a
feature, the data structure can be equally efficient. However, in general n-D
(n>1) cases, I there are significant differences among WCS (for GRID or rater
type) and WFS (for vector data type).
Hi Ben & Jon & all,
interesting to notice that your community requires coverage types beyond
what's listed in ISO 19123. This nicely coincides with the general
feeling that ISO 19123 (== OGC Abstract Spec Topic 6) deserves an update
- we might bring in your types on that occasion.
The best way for this probably would be if someone wrote a change
request to add a section about this your new coverage type. The WCS
group certainly is more than open to have a close look at such a