Wednesday, February 08, 2006
Generative Methodology Glass Bead Games
On using RDF to model web services
Doug,
again an excellent conversation about the issues.
as a general inquiry we find a mismatch between the SOA IM data object that "marshals" XML and some extended ontology where things might happen that no one has figured out yet... like the "discovery" of web services that an "intelligence" agent is looking for.
It is very difficult to grasp enough of the issues to make sense of where the industry is going. But there is a group that feels that a breakthrough is possible using a stratified approach to SOA. What is meant by stratified takes a while to understand, and yet there is adequate language to allow that understanding.
I am appreciative of the way you have stated your response.
Andrea was kind enough to make an extended response within the Protégé-OWL forum. And as I stated in the forum, this was very helpful to me. A very friendly effort on Andrea's part.
I am using Altova's rdf / owl editor... and now need to move some of the classes so that subsumption and other relationships exists... and decide in which case might a transitive or reflective property is appropriate. Altova does not (seem to) allow this move, so I will import into Protege OWL and see what happens. (The fact is that this import was not possible, and has lead to some extended discussion as to why and why there is a problem.)
I will try to complete this task and write up the issues to be posted for discussion on the protege-owl list. à [160]
Again, the general issue is how to make web services, which are XML and data objects, interoperable with RDF and in cases where this makes sense, OWL with its (as Andrea like to say) "semantic conditions".
Andrea's note is posted at: [158]
for later reference.
In the paper I will write over the next hours or days, I would like to quote you ….
"in the more general and common use of the term (in the software community), an "attribute" is very much like an rdf property. Owl properties are (specialized) rdf properties so the decision would depend on (1) how useful or important it is to distinguish between data types (text, integers, numbers, etc.) and objects in the range of things the property relates to and (2) how useful or important it will be to be able to operate on the information you create/store with automated systems that implement "reasoners" and related technologies.
which is very clear and useful to understanding how RDF and OWL might be used where its limitations are actually understood. In many cases, the limitation are perhaps a limiting factor on a larger project (ie the ontological modeling of web services).
Protege frames is also an option, as would be a KIF type "frame" based ontology. I would conjecture that frames with slots and fillers would be a more straightforward rendering of XML <--> data object persistence. But the confusion over the frames representation of human knowledge is just too profound at this point to work with. Maybe this will change?
PSP