- From: Jonathan Borden <jborden@mediaone.net>
- Date: Fri, 6 Apr 2001 12:35:47 -0400
- To: "Graham Klyne" <GK@ninebynine.org>
- Cc: "pat hayes" <phayes@ai.uwf.edu>, <www-rdf-logic@w3.org>
Graham Klyne wrote: > > > At 11:09 PM 4/3/01 -0400, Jonathan Borden wrote: > >I guess I'd argue at the phrase "fundamentally flawed" which implies (in > >english) that there is no hope to be fixed. I think RDF is > fixable and worth > >fixing. > > Jonathan, > > I think you and I are following similar paths here. hopefully we all are following similar paths, though when you look really closely at any river the particules of dirt appear to be following brownian motion :-) > > If I understand Pat correctly, his concern is, in part, that RDF > is either > too much or not enough. And to fix that requires some significant change > in the area where the RDF(S) specifications deal with "semantics". > Stripped down RDF is, to me, enough. There has been discussion about the utility of the triple model in its most simplistic form. My perspective has a strong emphasis on XML and the web. The utility of RDF as opposed to any old triple model that has been around for ? centuries, is precisely that the triples are composed of URIs. Argue about the problems with URIs (and I do think the language defining concepts such as "resource" and "entity" etc. needs to be cleaned up), however the simple fact is that the world has decided that URIs are something important. By the world I mean essentially the entire civilized world (not a closed world :-)), no one can deny the impact of the browser and HTTP. This makes URIs interesting things to describe. The question then is: is the stripped down RDF abstract syntax too simple to be useful? To me (as someone who has been interested in interpreting arbitrary or colloqial XML as RDF) RDF is precisely interesting because it is a _simplified_ XML "infoset", simplified from the XML infoset because it cares not about element order, nor whether an information item is an attribute or child element. > How would it go if we: > > (a) described the RDF core as an abstract syntactic framework > (the "graph") > for expressing semantic concepts -- but stopped short of claiming that it > was a semantic framework, and fine by me, but I'm just someone who implements stuff and then generally comments on what he has learned via implemention. as long as we are all talking about the same thing, we can have a useful conversation. > > (b) reviewed RDFS with a view to defining the simplest possible logical > framework that could be described as having a consistent semantics -- > something short of full FOL, but something on which FOL could be built > consistently, through the addition of additional formal elements. or simply replace RDFS as it currently stands with something like DAML+OIL, perhaps "DAML+OIL lite". All I want is a good way to write down ontologies in a way that is useful to machines (actually it should be useful to both machines and humsns). assuming the goal is to create an XML based "logic language" (to be very broad) a question I ask is: can this language be designed so it can be described by an RDF abstract syntax or does it need the full XML 1.0 syntax (or the subset of this that is the XML Infoset)? If so, there would be a practical benefit that triple stores could store DAML documents in a native fashion, rather than in an external storage system -- such a capability probably is useful for deployment and implementation purposes, and hence the utility of defining DAML in an RDF compatible fashion. > > ? > > I think (a) would continue to support the existing RDF applications, and > (b) could give us an opportunity to build a basic semantic framework that > is correct for developers and logicians. > > I appreciated your comments about the "non-Web experiment". I might > quibble with some details, but agree with your broad thrust. I would take all concepts of collections and reification out of the RDF 'model' or abstract syntax and define current collections as terms of physical syntax (e.g. what triples/URIs are to be generated as a result of parsing a document). this shouldn't cause problems for current applications which aren't really capable of dealing with reification or collections in a coherent fashion because people have not been able to agree on how these things are to be interpreted. -Jonathan
Received on Friday, 6 April 2001 12:35:14 UTC