RDF list vocabulary

From OWL
Jump to: navigation, search

This page is to document work related to Action 159, collecting the issue related to the possibility of allowing modelers to use the RDF list vocabulary in valid OWL2 DL. It is a work in progress.

The vocabulary

The RDF list vocabulary consists of:

  • rdf:List
  • rdf:first - this the most problematic one (see Boris' proposal)
  • rdf:next
  • rdf:nil

Benefits of allowing them

  • Allows more RDF to be consider valid OWL-DL

Cost of allowing them

  • Potential backwards compatibility issues
  • Tacit support of bad modeling practice
  • rdf:First DatatypeProperty versus ObjectProperty clashes

Implementor experience

Pellet

Mike Smith (Clark & Parsia) writes:

To summarize Pellet's "handling" of rdf:Lists... its not likely to work as expected if you try to use rdf:Lists in the ABox data and infer any results.

The Jena loader takes explicit action to ignore (discard) triples using the list vocabulary. There are ways some of the vocabulary can sneak into the KB, but not in a principled, intentional way.

The OWLAPI loader leans much more heavily on the OWLAPI implementation, so Matthew Horridge can probably tell you better what will happen. I believe that (at present) it will pun all the list vocabulary, so things might work out. Its more likely to go well if you only use it for object property or only data property since Pellet doesn't support object/data punning.

Use of the list vocabulary is certainly not something we test.

I'd like to give you more encouraging data, but that's the current situation as I understand it. I suspect Bijan was remembering that if he disabled species validation, Pellet wouldn't clearly break due to list vocabulary and seems to work ok under Protege 4 in the presence of it.

Technical issues

Proposed solutions

  • On the issue of Datatype/Object Property clashes - Have OWL2 be silent - clash pushes you to OWL2.
  • Have two subproperties of rdf:first, owl:firstLiteral, owl:firstObject per Boris' proposal
    • It is not clear that this solution provides the benefit claimed above - allowing more RDF graphs to be OWL DL - since, presumably, OWL DL would require using the new subproperties. (MikeSmith 05:53, 12 June 2008 (EDT))

Ontologies almost OWL-DL that use RDF lists