- From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
- Date: Wed, 19 Mar 2003 09:40:19 -0500 (EST)
- To: www-webont-wg@w3.org
Comments on Reference Overall Comment - Lots of work is still needed, particularly in sections 3, 4, and 5. I have not provided detailed comments here, as they would probably end up being longer than the sections themselves. General Comment - There are many places where pieces of RDF/XML syntax are described. These are not statements. It is also probably not a good idea to use ``fragment'' here also. I have thus used ``syntax'', but perhaps a different word would work better. 1.1 - See below for a suggested change in the status of Appendix B 1.2 - It is not "OWL Description Logic (DL)". Instead use "OWL DL" by itself. 1.2 - Be consistent in the capitalization of Description Logics. I suggest capitalizing both works. 1.2 - OWL DL and OWL Full do not support the same set of language constructs. For example, rdf:Alt is not allowed anywhere in OWL DL, but is allowed in OWL Full. 1.2 - e.g. -> e.g., ect. -> etc. Semantic -> Semantics [several other spelling mistakes around here] 1.3 - An OWL ontology is encoded and written as an RDF graph [RDF Concepts], which is in turn a set of RDF triples. As with any RDF graph, an OWL ontology graph can be written in many different syntactic forms (...). ... forms in writing OWL ontologies. However, ... by the RDF graph. ... resulting in the same RDF graph. ... the same RDF graph, consider the following RDF/XML syntax ... The following RDF/XML syntax ... encodes the same set of RDF triples, and ... 1.4 - ... Any RDF graph forms an OWL Full ontology. ... OWL Full ontologies can thus include ... 1.4 - Senantics -> Semantics 1.6 - see comment on Appendix B below 1.6 - ... lists many of the changes between ... 2. - OWL Ontologies Information in OWL is gathered into ontologies, which can then be stored as documents in the World Wide Web. One aspect of OWL, the importing of ontologies, depends on this ability to store OWL ontologies in the Web. 2.1 - An OWL ontology consists of optional ontology headers (generally at most one) plus any number of class axioms, property axioms, and facts about individuals. 2.1 - As with any RDF document, the information in an OWL ontology should be sub-elements of an rdf:RDF element. This enclosing element generally holds XML namespace and base declarations. As well, an OWL ontology document often starts with several entity declarations. For a typical example of this sort of information, see ... 2.2 - OWL built-in vocabulary The built-in vocabulary for OWL all comes from the OWL namespace, http://.... conventionally associated with the namespace name owl. It is recommended that ontologies not use names from this namespace except for the built-in vocabulary. OWL tool builders should feel free to signal a warning if other names from this namespace are used, but should otherwise continue as normal. 3 - I strongly suggest that the scare quotes be removed from class descriptions and class axioms and from other places in Section 3. 3.1 - The four latter kinds of class descriptions .... 3.1 - The first kind of class description is an instance of owl:Class, ... <owl:Class ...> NOTE: ... NOTE: ... owl:Class and rdfs:Class are the same there. 3.1 - Note: If one provides an RDF identifier for class descriptions of the latter 5 kinds, this has the extra effect of providing a way of referring to the class description by its name, i.e., as a class description of the first kind. See ... 3.1 - Remove ``Class descriptions .. following subsections.'' 3.1 - Two OWL class identifiers ... every OWL class is a subclass .... (For ....) 3.1.1 - ... of the enumeration kind ... This enables a class to be defined by ... For example, the following RDF/XML syntax defines ... 3.1.1 - The RDF/XML syntax <owl:Thing ...> refers to some individual. (Remember, ....) 3.1.2 - special kind of class description that defines the class of of all individuals that satisfy the restriction. A value restriction ... property. For example, we might want to refer to those individuals whose value of the property adjacentTo should be a Region, and then use this within a class axiom, perhaps even a class axiom for Region itself. Note that this is very different from rdfs:range, which applies to ... A cardinality restriction ... can take. ... the following general form: ... exactly one triple ... exactly one triple ... c.q. ??? 3.1.2 - If you are going to provide headings for owl:allValuesFrom, and other OWL syntax in 3.1.2, you should do it throughout the document. 3.1.2.1 - An owl:allValuesFrom restriction ... [This is currently both wrong and misleading, and needs a complete rewrite. Similarly for the other restrictions.] 3.2 - [This is very confusing.] 4 - [The characterisation of properties is not correct in this context. This makes much of this entire section both wrong and confusing. It would be much better to talk about how OWL property axioms as organic wholes, instead of breaking out sections on various properties and classes that are used in OWL property axioms.] 5. - [Individuals are not defined in RDF.] Information about individuals is given as in RDF. [This isn't very good either.] 6. - [RDF does not have attributes.] [more changes needed here] 7. - 8. - In OWL Full, owl:Class and rdfs:Class are identical, as opposed to OWL DL and OWL Lite where not all RDFS classes are OWL classes. Appendix B - I strongly suggest that owl.owl not be normative. Instead it should serve as an informative RDF Schema for part of the meaning of OWL Full. This also requires changing one of the paragraphs in Section 1.1 and part of Section 1.6.
Received on Wednesday, 19 March 2003 09:40:31 UTC