the difference between owl-s vs wsdl-s in semantic web services?

Question

i am acctually dealing with semantic web services, more specifically WS composition.

to add semantics in web services i find two paths (using owl-s or wsdl-s) ,so what is the limits of each solution?

Answer 1

In my opinion, both OWL-S and WSDL-S are equally obsolete. Both approaches were conceived at a time when heavy-weight service descriptions were thought to be the best way to build a web service architecture. Now, with the possible exception of some in-house corporate development teams, web services API's are generally based on fairly simple HTTP-based approaches, with JSON (usually) or XML (less often) as the payload. They often claim to use REST as a basis for defining the service, though in many cases it's far from clear that the term is being used correctly. By and large, these services are invoked through developers reading API documentation and writing code, rather than some cumbersome semi-automated process involving the processing of a service description language.

I'm not aware of any widely-used technique that can be used to describe the metadata of a modern web service of the form I describe above. There are approaches such as json-service , but I don't know how broadly they are used.

I think the truth is that having a rich service description as a first-class object never proved to be as useful as people expected, or hoped, it was going to be. It is especially true that we never achieved the flexible, dynamic composition of complex services from simpler component services envisaged in service choreography or agent-based web services . I think it's also true to say that describing what service would do upon invocation proved to be a much harder problem than the early research investigations anticipated, and premature attempts at standardisation actually made the problem worse not better.

If you're interested in web service composition, I strongly recommend you read some of the extensive research literature to get a sense of the scale of the effort that tried - and failed - to solve a problem that developers actually want to solve. Then ask yourself what problem for your users you are trying to address. Don't just invent some technology and then try to figure out what it might be useful for - that approach has been tried!

Answer 2

SWS have still long way to go. WSDL 2.0: RDF mapping W3C specification defines WSDL 2.0 - OWL mapping. The main problem is not how to describe the services in ontology, but what to do with type system. Linked Data is based on DL model (OWL2) which is quite different from XML Schema message descriptions. OWL 1/2 partially adopt XML datatypes, but you can't just use OWL2/XML serialization. One can use OWL/RDF messages directly (passing individuals with correspondent data), but there is no standard way to define this pattern in WSDL. The XML Schema is another brake here - it doesn't support multiple inheritance, so mapping between OWL / XML is not that easy (and to be fair in XML there no concept of inheritance at all).

IMHO the solution would be to create a new XML Schema language, keeping the type system, but changing a structural specification to something "Linked Data friendly". Then, provide new OWL serialization format based on say "XML Schema 2.0" specification, where all datatypes and individuals will be directly represented, so may be structurally represented in XML. Obviously there could be a lot of questions - how to represent OWL IRI in XML Schema QNames?

Kind regards,

D.