New model for describing services

[Guy Rixon]

On Tue, 9 May 2006, Roy Williams wrote:

> >
> > We're proposing a new approach to describing services which I refer
> > to as
> > "service with capabilities".  In this approach, a service record can
> > describe support for multiple service capabilities--such as
> > ConeSearch and
> > SkyNode--in the same record.
>
> In numerous talks on the Virtual Observatory over the years, I have
> told the registry model as "Dataset with Services".
> -- The Dataset is the provenanced, peer-reviewed data collection (for
> example 2MASS or SDSS).
> -- The Services are the URL endpoints and other technical information
> that allows you to access/query the Dataset (for example the 2MASS -
> Conesearch or the SDSS-SIAP).
>
> There is a citation system in the registry ("associated resource" I
> think) that connects registry resources together with named
> relationship types. The idea is that the Dataset cites the Services
> that deliver it, and the Services cite the Dataset they expose.
>
> (1) My first question is why we can't use this very flexible citation
> system for grouping multiple service capabilities? Why is a schema
> change necessary? Is there any plan to use the citation capability
> that we already have?

If we did it all by citation, we'd have to duplicate - in writing and in
parsing - all the common metadata for each service interface. Given that a
DAL service might have a primary interface, a couple of extended interfaces, a
VOStore interface for data staging, a delegation interface for security, plus
three VO Support Interfaces, that's a lot of registry records!

We do expect to use the associated-reource mechanism to link data-collection
records to services.

Consider also the case where a given data-collection is mirrored by several
archives. Each archive has a group of services related to this data-set and
related to each other: see the list above for examples. A client needs to uses
these in matched sets; no goof using the SkyNode interface from one archive
with the getAvailabiity from another! Therefore, it makes sense to group the
service endpionts in the registry.

> (2) Should I think of the new model as  "Dataset with Services that
> have Capabilities"? If so, what metadata is associated with the
> Service part, given that Dataset describes science and Capability
> describes technology?

Yes. The services have the Capability blocks describing access details. The
data-collection types have the details of the data structures. E.g. details of
queriable columns and tables are to be in the data-collection record and
details of query constaints (greatest box-size for SIAP etc.) in the service
record.

>
> (3) Or are you suggesting "Dataset with Capabilities"? In this case,
> I see little conceptual change except that the words have changed.

Don't think this matches our approach.

Given the set of concepts {service, application, data-set}, application and
(query on) data-set are nearer than service and application.

Guy Rixon 				        gtr at ast.cam.ac.uk
Institute of Astronomy   	                Tel: +44-1223-337542
Madingley Road, Cambridge, UK, CB3 0HA		Fax: +44-1223-337523