GINF roadmap (Re: SOAP, RDF and Semantic Web screenscraping...)

Dan Brickley wrote:
> ...
> (or xml-rpc for that matter). Given say a Java implementation that
> wrapped the above SOAP services, what extra conventions would we need
> before these informationally oriented RPC-ish services could be represented as
> RDF query servers?
> ...
> Dan
> 
> ps. I've a hunch Sergey's been building something similar to all this
> in GINF, though haven't quite figured out how GINF relates to XML-RPC,
> SOAP etc...


To answer some questions about GINF that popped up on the list, below is
a rough roadmap for GINF that I envisaged. SOAP-like wrapping is just
the first step...


1) Wrap Web services: so-called "canonical wrappers" allow communicating
with Web sevices using sets of logical statements, or RDF graphs (SOAP
has a similar goal, uses XML documents).

2) Specify mediators that allow interoperating between various services.
Mediators use mixes of (RDF-encoded) formal languages like state
machines, Datalog, etc. to perform stateful graph transformation.

3) Describe the interfaces of wrapped services, e.g. using a state
machine formalism.

4) Store wrapper interface descriptions and mediator specifications are
first-class objects in an RDF repository.

5) Develop tools for locating relevant services (by analyzing interface
descriptions). 

6) Develop tools for locating mediators that may be needed to
communicate with a relevant service (by analyzing mediator
specifications).

7) Develop tools for automatic maintenance of mediator specifications
when wrapper interfaces change.

8) Develop tools for automatic generation of mediators using the mass of
existing mediators (mappings between Web services).

9) Retire ;)
...

Recent conference papers that discuss some aspects of (1)-(3) are:

[1] ER'2000: http://www-db.stanford.edu/~melnik/pub/er2000.pdf
[2] DL'2000: http://www-db.stanford.edu/~melnik/pub/dl00.pdf

Received on Friday, 8 September 2000 21:36:13 UTC