W3C | Submissions

Team Comment on “OWLlink Protocol” Submission

W3C is pleased to receive the OWLlink Protocol Submission from the W3C Members:

  1. Clark & Parsia LLC (AC rep: Kendall Clark),
  2. Creative Commons (AC rep: Ben Adida),
  3. Daimler Chrysler Research and Technology (AC rep: Ingo Melzer),
  4. Free University of Bozen-Bolzano (AC rep: Enrico Franconi),
  5. German Research Center for Artificial Intelligence (DFKI) GmbH (AC rep: Massimo Romanelli),
  6. NTT DOCOMO (AC rep: Masaaki Yamamoto),
  7. Stanford University (AC rep: Natasha Noy),
  8. University of Aberdeen, Computing Science (AC rep: Jeff Pan),
  9. University of Manchester (AC rep: Sean Bechhofer), and
  10. Vrije Universiteit (AC rep: Frank van Harmelen).

Short Description

OWL reasoners are complex systems that, often, must be accessed via the Web. An HTTP based protocol is necessary for that purpose, in order to transfer possibly complex queries over the Web and get back query results. Clients should be able to upload ontologies, possibly remove some of the axioms or facts, and query subsumption hierarchies, class memberships, etc.

OWLlink’s Structural Specification defines a general, implementation-neutral protocol to access the functionalities of a reasoner acting as an (OWLlink) server. This general mechanism is defined in term of UML. Separate documents define bindings of this general protocol with different syntaxes that can be used to communicate with reasoners over HTTP. The submission includes a binding of the protocol with XML (based on the OWL/XML specification), with a functional-like syntax, and with S-Expressions (close to Common Lisp). Using one of these concrete protocol bindings clients can control and query reasoners using the terms defined in the general OWLlink Structure.

OWLlink’s Structural Specification is closely related to OWL 2. Conceptually, the OWLlink server relies on OWL 2 axioms, and the protocol covers such OWL 2 specific features like datatypes and datatype properties, usage of IRI-s, the notion of structural equivalence of OWL 2 ontologies, etc. An OWLlink server can declare OWL 2 profiles that it supports, or the name of the particular semantics it can use (i.e., RDF-based or Direct semantics). OWLlink also provides extension possibilities (although only the axiom retractions are part of this submission).

Clients can add individual axioms or load full ontologies to the OWLlink server; axioms may also be removed via the retraction extension. Queries to the client include all the typical queries one would expect from an OWL reasoner, like queries over the knowledge base (e.g., “is A a subclass of B”, retrieve a class hierarchy or property hierarchy, etc.), or queries over facts (“is indiviual X of class A?”, are two individuals equal or different, etc.).

In summary, OWLlink may play a major role in the management and the usage of OWL 2 reasoners; as such, it may become an important element in the deployment and usage of such reasoners on the Semantic Web.

A technical comment: the HTTP protocol is using exclusively HTTP POST, including for operations like queries for which a HTTP GET might be more natural. It would be good if the designers of OWLlink also considered using HTTP GET and, in general, RESTful protocol design wherever appropriate for newer versions.

Next Steps

At this moment, W3C does not have any active group on OWL and has no plans to create one in the near future. However, W3C will be pleased to consider this submission in the framing of any future activity involved in the development of OWL.

Ivan Herman, Semantic Web Activity Lead <ivan@w3.org>,
$Date: 2010/07/21 15:34:40 $