Re: does DAWG actually have time to do WSDL?

On Mar 21, 2005, at 4:11 PM, Seaborne, Andy wrote:

> Bijan Parsia wrote:
>> On Mar 21, 2005, at 2:21 PM, Dan Connolly wrote:
>>> On Mon, 2005-03-21 at 13:27 -0500, Bijan Parsia wrote:
>>>
>>>> On Mar 21, 2005, at 1:12 PM, Dan Connolly wrote:
[snip]
>> One thing I may not have mentioned is that it make useful service 
>> composition very difficult to impossible.
>
> Service composition is a large, open ended area and I don't have a 
> characterization of the problem as applied to SPARQL.

There's at least two more specific applications. One is to use SPARQL 
queries in the specification of preconditions in OWL-S a la how 
conjunctions of SWRL atoms are being used now. That's independant of 
protocol. The second is that I might want to compose invocations of a 
SPARQL service with other services, for example, if I have an owl-s 
description of  SPARQL server that tells me that it stores information 
about flights using a certain RDFS schema, I may want to generate 
certain queries that will then be input to a booking service.

There are more.

> For example, it isn't automatically a need for an XML syntax.  We do 
> service composition but by pulling out the abstraction and combining 
> conceptualized services - that means abstracting away from either XML 
> or human readable synatx and the presence of an XML form does not aim 
> the process significantly.

The advantage is in the extra type information, not the XML surface 
syntax. The use of XML Schema in the WSDL abstract interface is 
primarily to allow for static typing. xs:string doesn't distiguish a 
sparql service very well.

On the wire, it can be whatever you want.

In other words, in WSDL, the abstracting away is done by means of XML 
Schema (built in). There is a nice advantage that there is a common, 
indeed standard, concrete version of that abstraction suitable for the 
wire. We get that for free once we get the schema type.

Cheers,
Bijan.

Received on Monday, 21 March 2005 21:21:32 UTC