[esip-semanticweb] Interface for Tool Match query service

[Wilson, Brian D (335G)]

I second the vote (with Eric) that we should use a query URL syntax to describe
the ToolMatch service.

Strict REST services are for performing CRUD operations on a resource.  One can
hierarchically browse resources by following directory paths, but even if multiple
paths are set up one will only have a few "views" of the resources (here datasets
and tools).

I suggest that one could use our extended OpenSearch protocol to put up the
tool query service, with dataset ID as the primary parameter, but also text
keywords and custom keywords for all of the other "facets" folks want to
use to search for tools (OS, maps, etc.).

The query response would then be the usual Atom feed with tool metadata
inserted in each entry.

The service casting standard already handles metadata for all kinds of services,
including GUI's intended for humans.  With a few extensions, one could handle
the more general category of "tools" and fit the metadata into a cast.

If the ToolMatch back-end is RDF and an ontology, OpenSearch could still be the
standard "simple" interface, hiding the complications of SPARQL and the fact that
RDF is being used at all (just as OpenSearch currently hides an SQL or solr query).

Besides, in the casting work we are already heading toward having a mapping
of the services & collection metadata casts into the RDF world.  There is no
reason why we can't have complete interoperability between XML representations
of metadata and equivalent (or more general) RDF representations. 

The same goes for ISO metadata.  In a not so distant future, I at least hope that
we will have interoperable representations of ISO in both XML and RDF, so that
modular pieces of ISO metadata (expressed in XML, RDF, JSON, whatever) can
be embedded in cast entries, <link>'ed to from cast entries, manipulated &
displayed in web GUI's, and authored & added to in web GUI's.

Moreover, some ISO modules, like that for "lineage", need to interoperate with
other metadata standards, like the Open Provenance Model (OPM) or the
coming W3C PROV.

It's good to have lofty goals.  It should only take a few years.   :-)

 -- Brian