- From: Graham Klyne <GK@NineByNine.org>
- Date: Thu, 28 Nov 2002 15:57:53 +0000
- To: RDF core WG <w3c-rdfcore-wg@w3.org>, Frank Manola <fmanola@mitre.org>
RDF Primer W3C Working Draft 11 November 2002 This version: http://www.w3.org/TR/2002/WD-rdf-primer-20021111/ I think this is looking very good. I haven't done a detailed read through: I started looking for overlaps with RDF Concepts. Those overlaps that I noticed (e.g. URIs, RDF graph model, structured information) had such different forms of presentation that I think they really stand separately as extended explanations, with very little real overlap. [RDF graph model: cross-ref from concepts? Likewise simple facts -> structured data?] Buried in section 3.2 (syntax) is an important discussion of anyone being able to add RDF assertions about any resource. Maybe this should be more prominent? Section 4: built-in *types* should be *classes*?? Section 4.3.1: maybe needs to split into two sections; one to deal with breaking up n-ary relations, and another to introduce rdf:value? General comment, not specifically Primer: the description of rdf:value is fine, but how does it relate to a normative specification? What can we say formally about rdf:value? What formal semantics (interpretation) allows us to make inferences like: my:cat rdf:type ex:DomesticCat . my:cat ex:weight _:x . _:x rdf:value "15" . _:x ex:unit ex:Kilogram => my:cat rdf:type ex:Obese . but NOT: my:cat rdf:type ex:DomesticCat . my:cat ex:weight _:x . _:x rdf:value "15" . _:x ex:unit ex:Pound => my:cat rdf:type ex:Obese . ? My point here is if we are to encourage such usage of rdf:value, then there ought to be some normative description to back up such usage. Section 4.3.2, given the links to XML schema datatypes we now have, and the subtleties (i.e. non-obviousness from a typical programmers' PoV) of the alternative approach of using rdf:type for true/false values, and the distinction of meanings viewed from an open-world perspective, I wonder if this section really belongs here? Section 5, para 2: (RDF schema) mechanisms to *describe* rather than *specify*? Section 5.1, para 2, typo: "different kins of moFor" Is it worth noting, in passing, that it is not necessary to define a schema to use new vocab, even if it's recommended practice to do so for the purposes of documentation and consistency checking? 5.3: very good! 5.5: "specifying that two different classes, defined in separate schemas, actually represent the same concept." Or have the same members? (I think this is a subtle difference between the RDF view of a class and the DAML/Description Logics view.) 6. Examples. I've only skimmed these, so I may be missing something. I think it would be useful to have an example that shows some simple inference in action. I have come to the view that it is the ability to have standard tools do inference -- even very simple inference, and the ease of doing so, that sets RDF apart from other network data formats like XML. I'm thinking of maybe something of DanC's work relating to travel schedules might work. A simple example I like is DanC's tools to create graphs from arbitrary RDF: at heart, a simple cwm inference rule defines the nodes to be graphed. Section 8: does it make sense for a non-normative document to have normative references? #g ------------------- Graham Klyne <GK@NineByNine.org>
Received on Thursday, 28 November 2002 12:03:44 UTC