From: Ian Horrocks (horrocks@cs.man.ac.uk)
Date: 03/03/01
On March 3, Dan Brickley writes: > On Sat, 3 Mar 2001, Ian Horrocks wrote: > > > On March 2, Dan Connolly writes: > > > Dan Brickley wrote: > > > > This is a followup to some hallway and lunchtable conversations about > > > > DAML datatyping and the work of the Dublin Core Metadata Initiative. > > > > > > > > My understanding of the forthcoming revision of DAML+OIL+DT(*) is that > > > > we say all properties are either of the kind that point to resources, or > > > > of the kind that point to concrete datatypes, strings structured as per > > > > XML Schema part 2. > > > > > > I don't think that's quite right; I'm not confident I know where > > > the latest draft is, but my understanding is: DAML+OIL+DT doesn't > > > say that all properties are either black or white; it just doesn't > > > tell you the semantics of the grey ones. > > > > This is exactly the case: DAML+OIL+DT defines two subclasses of > > Property, AbstractProperty and DatatypeProperty. > > So I can re-assure the DC folks that DAML processors won't be throwing > exceptions or refusing to load DC-based data structures because of this? Well, we obviously can't guarantee what attitude any given DAML+OIL processor will take to arbitrary RDF, but there is no reason for it to barf - it could just ignore those RDF triples to which DAML+OIL does not assign any semantics. > (context: this thread was occasioned by a lunchtable remark from TimBL > along lines that "I'm not sure DC will survive DAML"...). > > An alternate strategy for the 'grey ones' is to wait for RDF-logic rules > machinery to annotate things like dc:creator with conclusions that can > be drawn in different contexts. So while the basic definition of > dc:creator is very loose and grey, additional claims made by DCMI might > say things like: when dc:creator points to a literal, it is the name of > an un-named resource of type dc2:Agent. While I'm not sure this is the > best route for DC, it makes sense given the history: we adopted this > loose modelling stratgy for DC because RDFS barely existed when DC > wanted to define a representation in the RDF model. I agree with Dan C. that this style of modelling is "broken". Ian
This archive was generated by hypermail 2.1.4 : 04/02/02 EST