- From: Jonathan Borden <jonathan@openhealth.org>
- Date: Fri, 13 Jun 2003 10:13:12 -0400
- To: "Paul Prescod" <paul@prescod.net>, "Tim Bray" <tbray@textuality.com>
- Cc: "WWW-Tag" <www-tag@w3.org>
Paul Prescod wrote: > > My last message on the topic. :-)) never say never again. :-)) > > Tim Bray wrote: > > ... > > I dislike it because it treats nature and purpose asymmetrically. > Nature and purpose can be syntactically symmetric (for certain syntaxes) and semantically asymmetric. > > A thing has a nature (which is to say type). A VCR is a VCR. > > You are a man. Pretty much everyone would agree. That's just your > nature. Nature is a property solely of you. On the other hand, purpose > is a relationship between you and the person "using" you. agreed. > > Your purpose to Antarti.ca is as a founder. > Your purpose to Lauren is as a husband. > > Now consider these assertions: > > Tim is a man. > Tim is a husband. > Tim is a founder. > Lauren is a wife. > Antartica is a company. > > The first statement is fine but all of the rest discard important > information. For all the computer knows, you are the husband of > Antartica or the founder of Lauren. Well, not quite, because the computer also knows that Antartica as a company is not in the class of things that may have a husband, but you can construct another example that demonstrates the appropriate problem e.g. a group of friends each of whom is married to _one_ other in the group. > > >> a) of Dan's point that your declarations do not bind appropriately to > >> the actual resource being described > > > > > > I disagree. I claim that rfc2396.txt has a nature (.../text/plain) and > > a purpose (...#normative-reference). > > It is a normative reference to some specs and an informative reference > to other specs. If you don't make the link explicitly then how will it > get made? Dan raised this earlier and I don't think you answered > directly. I think that Dan's model makes the purpose type information a > little harder to find but the other model really discards important > information. > > On the other hand, you are right that RFC2396 is ALWAYS text/plain, > wherever it is referenced. RFC2396 need not _always_ be text/plain (e.g. content-negotiation) but nonetheless RFC2396 is a member of the class of things that has a text/plain representation. > > And by the way, should rddl:nature map to rdf:type? I really don't know > what rdf:type is for if not to say that "RFC2396" is a "text document" > or "foo.css" is a "CSS document." When does one use which of > RDF/RDFS/OWL's various type mechanisms? And when does one invent ones own? > I've modelled rddl:nature as rdf:type. This says that "RFC2396 is a member of the class of things that is a text document". RFC2396 might also be in the class of things that is an RFC, an HTML document, etc. etc. -- this is to say that an individual may have any number of rdf:types (the individual may belong to any number of classes). RDFS/OWL's 'type mechanisms' are simply class definitions that might allow a computer to _infer_ that an individual is a member of a particular class, just as XML Schema's type mechanism might allow a computer to infer that the string "123" belongs to the class of strings that has an integer representation. I'd also say that rddl:purpose can be used to represent an rdf:Property (a property is attached to an individual). Jonathan Jonathan
Received on Friday, 13 June 2003 10:13:27 UTC