The tentative approach we are considering for WCS REST is to consider each dataset to be
a resource. A subsetting operation on a dataset is a "view" of a resource, not
a separate resource. The view/subset would be specified by query parameters. What are the
valid query parameters would be probably the main thing to clarify.
Whether this ends up with anything very different than what we already have
remains to be seen. It may be that this just drives things like HTTP response
The intention is to experiment with an alternative specification of equivalent
functionality that would be much simpler than the existing one. Ideally, it
might turn out to be a profile.
Jon Blower wrote:
So it may simply be that extraction-from-dataset-oriented services can
never be truly RESTful?
I fear so.
Having said that, if we take a feature-centric approach then one might
consider that a dataset (or "data granule" in NERC Data Grid parlance)
contains a finite number of features, each with its own id. It might
be possible to have a RESTful approach to getting these features, but
I bet it wouldn't take long for the REST model to break down. Is it
really possible to identify in advance all the possible features that
might be extracted from a dataset? I doubt it, even for simple cases
such as a satellite image.