| [higgins-dev] LDAP "person" sample ontology based on Higgins ontology |
Mark, Are you on the higgins-dev mailing list? Just wondering, no big deal, Paul was telling us today about a group being created where everybody can collaborate on identity schemas so I'll send this out to that group when it gets created but Jim and I wanted to get your feedback ASAP on this work. All, Attached is a sample of how we might represented an LDAP "person" (2.5.6.6) class in OWL by extending the Higgins ontology. BTW, I can't bring it up in Protégé because it throws an exception. It appears well formed to me but can someone with a different OWL editor try it to see if it works and declares the offending construct? I adopted several conventions from Sebastian Dietzold's work at the University of Leipzig, again, described here (www.semanticscripting.org/SFSW2005/papers/Dietzold-RDF_Models_from_LDAP.pdf - Generating RDF Models from LDAP Directories). Namely: 1. Labeling each class and property with an english string representing the item as it is commonly known in LDAP. 2. Defining an equivalent item named by the respective OID allocated for it in LDAP. 3. Creating a comment referencing the RFC where each item is described. 4. Single-valued attributes use a cardinality restriction (<=1) for the class "top". 5. Some syntax mappings to XML schema datatypes. This sample doesn't include any required attributes that employ item #4 but there's an example included at the bottom in comments. There are several items that are conveyed within LDAP Schema that I do not currently know how or if they could be mapped into OWL. What seems apparent is that is possible for users to get some value out of Higgins even if some of these cannot be conveyed. The underlying context provider and ultimately the back end will obviously still be free to deny anything that isn't considered kosher. Obviously, we'd prefer that OWL be expressive enough to convey everything. That said, here are the items I've come up with so far: 1. Abstract classes - is there a way in OWL DL to say that a class cannot have an actual individual declared to be of that class? 2. Auxilliary classes - we can define these classes but can we restrict their use how we would like to? 3. Restricting classes to only MUSTs and MAYs. 4. Matching rules - how would we express these? Do we need to? There are some items (that I know about ... plus all those pesky ones I haven't thought of yet) that are TO DO: 1. Complete examples of enough LDAP Schema elements to cover all the cases we need to handle. 2. Map all syntaxes to either literal or complex types. 3. Write code to dynamically convert LDAP Schema into an Higgins based OWL ontology. Tom
Attachment:
HigginsLDAP0.1.1.owl
Description: Binary data