W3C home > Mailing lists > Public > www-rdf-interest@w3.org > November 1999

Re: Metadata, XML, and RDF

From: Sankar Virdhagriswaran <sv@crystaliz.com>
Date: Mon, 15 Nov 1999 20:35:12 -0500
Message-Id: <199911160136.UAA02703@hunchuen.crystaliz.com>
To: Stefan Decker <stefan@db.stanford.edu>, www-rdf-interest@w3.org
I agree with most of what you are saying.

You are also correct in saying that it is more a frame system than predicate
calculus based system. But, I thought it did a decent job of unifying both
of these perspectives. This is also not new. All the commercial expert
system tools of
circa 1984 did this (ART, KEE, etc.).

Finally, when I mentioned practical applications, I was talking historically
and 'commercial' applications. I am not aware of any commercial efforts that
use RDF for reasoning purposes in a major way  (I am aware of open

thanks for the clarifications.

>From: Stefan Decker <stefan@DB.Stanford.EDU>
>To: www-rdf-interest@w3.org
>Subject: Re: Metadata, XML, and RDF
>Date: Mon, Nov 15, 1999, 8:07 PM

>>I am sure others would have different views on this, but the way I have
>>thought about this is to think of RDF as a conceptual modeling language and
>>XML-Schema as a logical schema language. RDF comes from the knowledge
>>representation community and hence worries a lot about semantic
>>representation and reasoning through them.
>I wish that would be true. Actually, if you want to map RDF to communities,
>it was initially developed as a metadata standard to represent Dublin Core 
>and to supercede the PICS standard.
>However, as soon as one starts to represent information in a general,
>declarative and reusable way,  Knowledge Representation techniques are
>indeed applicable.
>And the RDF-developers did a good job, many Knowledge Representation
>Formalisms are mappable to RDF, and RDF has the chance to be the
>first widely used Knowledge Representation Language.
>AI-folks are only partially recognizing this...
>>The most practical application of
>>this sort of stuff has been in conceptual modeling languages such as UML
>>(which is why reading the comparison report is interesting) and in expert
>>systems (or limited versions of these).
>I disagree. UML is a Software Engineering Methodology - which involves
>also modelling. However,  the most practical application right now is to 
>represent metadata
>about web resources and to built search engines for this.
>Indeed RDF is now used in several DigLib-projects to represent vocabularies
>used for classification of documents.
>And here RDF has to prove its usefulness.
>>XML-Schema is verbose and voluminous, but is driven by the need to support
>>exchange of document + ORDBMS data and with the need to support different
>>namespaces for schema. There are a lot of practical applications for this
>>sort of stuff.
>Sure. However, it is document-centered. It defines what is allowed for a 
>>RDF on the other hand worries about a new way modeling resources from which
>>'automated' agents could reason. One way to think about RDF is Prolog
>>clauses. The analog of the Prolog unification (i.e., inferencing) is what
>>automated agents that process RDF will have to write. The choice of syntax
>>in RDF makes it difficult for folks who are used to the style of mark-up in
>>XML and HTML.
>Not prolog clauses, but maybe facts. That is indeed the way RDF is used in 
>However, RDF is more a frame-system, talking about classes, instances, 
>and so on. BTW: Unification maybe regarding as Inferencing, but in this case
>it is really trivial....
>RDF can be imbedded in several logical formalism, one is Horn clauses,
>as e.g. implemented in one particular (not necessarily the best for RDF)
>way in Prolog.
>However, this is not the first aim for RDF, but using it for reasoning is 
>for sure
>something that can be done with it.
>>As such, representing your information in RDF allows for a lot of power and
>>flexibility. As can be seen from recent traffic on this group, one can start
>>with querying and go all the way up to inferencing. Furthermore, RDF Schema
>>makes expression of object oriented models fairly straight forward.
>         Stefan
Received on Monday, 15 November 1999 20:35:03 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:44:21 UTC