Re: The semantic web

About nesting - you might like to compare the following two files (XML and RDF)
for an online journal holdings to see the bloat factor:

http://www2.harcourt-international.com/ideal/metadata/xml/aama.xml
http://www2.harcourt-international.com/ideal/metadata/rdf/aama.rdf

Any comments about the RDF appreciated.

Tony





Pierre-Antoine CHAMPIN <pachampi@caramail.com> on 11/04/2000 16:14:26

To:   Tom Van Eetvelde <tom.van_eetvelde@alcatel.be>, "www-rdf-interest@w3.org"
      <www-rdf-interest@w3.org>, Pierre Maraninchi <penguino@caramail.com>,
      frankh@cs.vu.nl, Pierre Maraninchi <penguino@caramail.com>,
      dieter.fensel@aifb.uni-karlsruhe.de
cc:    (bcc: Tony Hammond/AP/LDN/HARCOURT)

Subject:  Re: The semantic web

> I have read the article "Practical Knowledge
Representation for the Web" (
>
http://www.cs.vu.nl/~frankh/postscript/IJCAI99-III.html#Calvanese:98AAAI
) and to my surprise, RDF
> is depicted as an immature and disappointing language for
representing semantics. This is in clear
> contrast with the promise of RDF as the language to make
the Semantic Web come true.
> The authors could be biased as they come from the AI
world, but nevertheless, their arguments seem
> well founded to me.

thanks for the link; very interesting article

> Does anyone have comments on this article?

I do :)

- about nesting (3.3)
I think the author are quite unfair : RDF CAN do nesting.
Sure it is more verbose than plain XML,  and hence less
readable - is this what they mean when they write that
nesting is not expressible "in a natural way" ?
Furthermore, this is a syntactical issue, and I think the
whole RDF community agrees on the necessity of a simplified
syntax. (Personnaly, I'm even much confident in an
attribute-based syntax - in the XLink fashion - allowing to
interpret any XML tags as RDF).

- about RDF beeing property-centric (4.1)
ah ! This is always the itchy part for people used to
object oriented models. I believe this is not a mistake,
though : once a property has been defined, any new schema
can use it, and therefore be (partially) understood by
anyone understanding that property.This is the whole point
- and this is possible because the domain of a property can
ALWAYS be extended (since it is not bound to be unique).
About translating Ontobtoker ontology into RDFS (4.2) ,
prefixing class-names to property-names is not the best
solution : using a different namespace for each class is
much more elegant.

- about inferences (4.3)
here is the most unfair point, IMHO : the kind of inference
proposed here is straightforward in RDFS with
subPropertyOf.

 Anyone seeing anything I forgot ?

  Pierre-Antoine

_______________________________________________________
Vendez tout... aux enchères - http://www.caraplazza.com