- From: Drew McDermott <drew.mcdermott@yale.edu>
- Date: Fri, 25 May 2001 11:18:03 -0400 (EDT)
- To: www-rdf-logic@w3.org
I am getting confused about what the symbol "RDF" encompasses. Jonathan Borden proposes an "abstract syntax" for RDF. (http://www.openhealth.org/RDF/RDFAbstractSyntax.html#RDF-MS) I am not sure what makes it an abstract syntax, but that's not important; the important thing is that it defines a language that violates the usual rules about RDF documents being collections of triples. That's because he proposes that they be considered collections of 6-tuples, one of whose components is a boolean saying whether the 6-tuple should be considered to be asserted or used in some other way. I think this is a big improvement on the triples model. Similarly, Jos DeRoo proposes treating embedded RDF in a special way: I think it's like how 'subexpressions' work, or how would you call the premis, conclusion, rule (or head, body, clause in Prolog)? So I would say option 2 (using log:quote to instruct an RDF/XML parser, but an RDF/N3 parser could just 'recurse' into those 'contexts' (at least that's how we did in http://www.agfa.com/w3c/euler/Euler.java)) Contrast this with Frank Manola's remark a few days ago: Correct me if I'm wrong, but an example of the problem would be that you've decided to add something like a NOT relation and have it mean that certain domain triples (which you also provide in the same "place" and which you reference in instances of the NOT relation) are not true. A processor that understands the NOT "language" of course does the appropriate thing, but an ordinary RDF processor sees what it thinks are asserted domain triples, doesn't understand what NOT means, and gets a meaning opposite to what is meant. I read Jos proposing one set of ground rules for extending RDF, Jonathan and Frank another. The DeRoo approach is to use quotation to hide nonstandard extensions of RDF. That is, an RDF processor that isn't in on the trick will see an uninterpreted string and leave it alone; while a processor that knows what "log:implies" means will de-quote the string and interpret it as if it had never been quoted at all. The Borden/Manola approach is to just let uninformed processors misinterpret RDF documents that go beyond the vanilla model. I personally favor the second approach, on the grounds that it's too early in the development of web metadata to wire in a convention that requires us to pretend to quote things all over the place. Ten years from now it will be very hard to explain to novices why the quotes are where they are. Anyway, if we're going to "let a hundred flowers bloom" when it comes to RDF extensions, then I've got to go water my garden. -- Drew McDermott
Received on Friday, 25 May 2001 11:18:10 UTC