- From: Jos De_Roo <jos.deroo.jd@belgium.agfa.com>
- Date: Tue, 20 Aug 2002 00:57:20 +0200
- To: pat hayes <phayes@ai.uwf.edu>
- Cc: Guus Schreiber <schreiber@swi.psy.uva.nl>, www-webont-wg@w3.org
> There is a very incomplete draft of the document I promised to write at > > http://www.coginst.uwf.edu/users/phayes/RDFS2OWL.html > > I will be updating it over the next few days at irregular intervals, > but the basic outline shouldnt change much. > > Any feedback would be welcome. There are a couple of places, > indicated by comments, where Im not entirely sure that the MT is > right, but I have followed Peter's semantics in any case. [...] > In a sharp contrast to RDF(S), classes and properties are > not first-class entities in OWL: they cannot be in the > domain and range of properties, or be contained in classes. > In fact the OWL universe is even more restricted than figure > 4 illustrates, since OWL properties (which may apply only to > things) may have values which are things or datatype values, > but not both, giving a parallel distinction between two kinds > of property. this is simply a different design than RDF(S) and the SW design as elaborated in http://www.w3.org/DesignIssues so please let's first of all resolve that "sharp contrast" [...] > 3. Skirting paradox: Inconsistencies in OWL/RDF > @@ discussion of the 'Peter paradox'. Different views on > how to treat it, and whether it should be considered to be > pathological. Semantic analysis in GHOWL. Imposing extra > 'global' syntax restrictions to enforce tree-structured graphs. > Conclusion: using a recursive syntax certainly avoids such > nasties, but they are not fatal; if one wishes to stay in the > RDF world, it is possible without incurring paradoxes. @@ that is definitely good news > 5. Injecting syntax into semantics. > While the semantic conditions described above faithfully > capture the appropriate meanings for the OWL abstract > syntax when rendered into RDF, it is clearly less elegant > to have conditions stated in terms of entire subgraphs rather > than on a triple-by-triple basis. For example, this requires > inference engines to 'assemble' entire subgraphs in order to > draw conclusions, and it provides no meaning to well-formed > RDF graphs which have incomplete OWL subgraphs, eg incomplete > lists or partially formed restriction triplets. Moreover, > ghowl-entailment cannot be defined in terms of closure rules > and rdf-entailment. I'm not completely understanding this paragraph and in particular it's last sentence (as a matter of implementation fact, we do *both* assembly of function trees and their triple assertions and when the former is not giving unifiers the latter mostly do) -- , Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/
Received on Monday, 19 August 2002 18:58:02 UTC