- From: Jonathan Borden <jonathan@openhealth.org>
- Date: Thu, 11 Apr 2002 09:27:26 -0400
- To: "TAG" <www-tag@w3.org>
Tim Bray wrote: > > I'd like to thank Stuart for all the spadework on this issue, which > I realize now that I didn't understand before. I'm not sure that I > understand it now, but I'm going to suggest another approach based > on Stuart's work that I think comes out a little cleaner: > > <div ID="schematron" class="resource"> > <h3>7.7 Schematron</h3> > <p>A <a href="rddl.sch">Schematron Schema</a> for RDDL. </p> > </div> > > <rdf:description rdf:about="rddl.sch"> > <rddl:prose rdf:resource="#schematron" /> > <rddl:purpose > rdf:resource="http://www.rddl.org/purposes#schema-validation" /> > <rddl:nature > rdf:resource="http://www.ascc.net/xml/schematron" /> > </rdf:description> This approach has an easy to read syntax yet I am concerned about _this particular_ RDF semantics and the applied relationship to a "namespace". RDF is a very precise and at the same time sometimes unforgiving tool. The primary distinction between RDF and any old XML is that the meaning, the formal semantics, of an RDF document is formally specified, thanks to the excellent work of Pat Hayes and the rest of the RDFCore WG. What this means is that you _are not_ allowed to decide what the RDF means, much as you might when writing down any other XML when devising a preferred syntax. Like it or not, RDF does have an unambigouos formal meaning aka a 'model theory'. The downside of this is that you can't just simply devise an RDF compliant syntax because it looks good, or is readable by a person. So while the syntax of the RDF that you've written looks terrific, the semantics are way off what a namespace definition ought to be. My comments are specific to _this specific_ RDF + XHTML, not the concept of RDF + XHMTL in general. What you get when parsing such a document is a collection of RDF resources whose names are the URIs of the namespace related documents. Fine, except... To my understanding a namespace is a collection of terms, not arbitrary resources. RDDL was carefully designed to link each term to a related resource, the RDDL purpose of the link is the name of the arc traveling from term defined by the rddl:resource id, to the related resource. The rddl:nature is interpreted as the rdf:type of the related resource. To me, this is the proper RDF representation of an XML namespace, and the RDF+XHTML syntax needs to be both readable, and to properly represent this relationship to a machine. I am not sure what to think of the names in the namespace being defined by IDs on <di v> elements, while the semantics of the namespace being defined by RDF. This feels wrong to me, as if you are saying that the terms in the namespace denote pieces of XHTML. You see when using anything other than RDF, what the terms in a namespace _denote_ in a formal sense, are either undefined or defined by the specification, in RDF, having a model theory, the denotation is unabiguous and predefined. If you think that's just a bunch of mumbo jumbo then you really ouight not be using RDF. > > ...but there's *lots* of room for argument about these details. > > I do think however that the way I've proposed structuring > the RDF description is the way to go. And (not a surprise > when you think about it) it's easier to read & understand > than the XLink formulation. -Tim Some formulations make more sense to people, some to machines. What we need is a formulation that makes sense to both people and machines, but particularly 1) that names in the namespace are defined using RDF terminology/vocabularly definition -- and desirable --2) that browsers understand these terms e.g. as fragments pointing to the prose descriptions. e.g. the way a browser handles URI '#' fragid. This formulation handles the latter (2) part (which is certainly desirable) but does not handle the former (1) part which I think is mandatory. The other issue is that when I think of using RDF to define namespaces, I don't think of predefining a couple of predicates in the RDDL namespace, rather really using RDF to define namespaces using arbitrary predicates. The advantage of RDF is that it is incredibly powerful. The disadvantage of RDF is that it is incredibly powerful. One of the reasons I think RDDL is that it constrains what you can say, which is great for XML folks that want to create namespaces without getting into the details of formal semantics. I've started to convert the RDDL spec itself into a blend of XHTML + RDF http://www.rddl.org/rddl-rdf.html -- you'll need to "view source" and look for the RDF statements intermixed with the XHTML. This RDF is intended to have exactly the same semantics as the RDDL XLink is intended to convey. If you don;t like this syntax, we'll need to do more work (it is not totally clean to intermix RDF + XHTML). Jonathan
Received on Wednesday, 10 April 2002 21:18:35 UTC