Re: RDF, RDFS and DAML+OIL benifits over XML (fwd)

Hi Libby,


On Wed, 6 Nov 2002, Libby Miller wrote:

> 
> Thanks Pedro - this gives me a chance to try out a frequently asked
> question answer.

:) Please see my comments in-line.

> My personal feelings are represented in [3] - if you ever need to reuse
> or share - use RDF over plain XML.

Ok. Suppose that I can achieve greater flexibility in managing RDF/RDFS
model evolution (version, reuse, share, etc) than with OO-modeling+XML.
Eventually, in XML case, if something is changed at the bottom level -
knowledge model description - it should be required to change a lot of
things: model, XML translation, DTD/Schema, parsers, etc. But, IYHO how
is this handle in RDF/RDFS approach? Is it about RDF triples information
model, RDFS OO modeling concepts, both, or something else?

> 
> Q: RDF and XML - which one do I choose?
> 
> A: XML is a very powerful language for describing heirarchical things.
> It defines what goes into a document. For 'known target' interactions
> XML works very well.
>
> (snip) 
> 
> For example, if you or I read the above we might think that '<name>' was
> related to the tag above it '<person>' in that it was an actual
> person's given name; and we would realize that a W3Cprofiledate was a
> specific type of iso8601date. But structurally, and
> syntactically, they look very similar in XML even though the
> relationship between each pair of elements means something completely
> different. Unless machines pre-co-ordinate, machine understanding of
> these relationships is very hard because of the implicit, not
> explicit semantics in ordinary XML. RDF, through schemas and extensions
> like DAML+oil or OWL can provide machine-readable semantics, even when
> there is no pre-coordination.
> 

Yes. XML is document-centric and relies on parser hard-coded information,
meaning that it is machine readable, regarding its contents, eventually
with document validation through distributed/centralize DTD or Schema. So,
elements semantics (or if you like theirs metadata) should be in the
parsing rules body, but it can be addressed through specific
elements/attributes that provides the metadata about the element itself.  
But, in the end it remains a rigid structure and still reveals a somewhat
implicit metadata schema. It is clearly a rigid structure that mirrors an
also document rigid structure, revealing a syntactic-oriented approach.

So, the RDF/RDFS advantage is that it moves an layer up the OO concepts,
and implicitly reveals all OO common advantages: <person> is a class (e.g.  
DAML) and <name> is a class property, and from that I can get more
information than XML <person> element (flat model representation).


> There is also a difference in the perspectives of RDF and XML: XML is
> about documents; RDF is about the world. XML is about the number of
> elements of certain types, their attributes and ordering, and the sorts
> of text that can appear in datatypes, while RDF is about things in the
> world -  people who have names, create documents, have friends. RDF
> is about real things in the world not the documents that describe them.
> 

Yes, I acknowledge that context difference (thing class). That is clear
for me.

I'll take a closer look at the links that you provide seeking for more 
understanding.

Thanks for the reply.

Regards,

-- 
Pedro

passis@dee.isep.ipp.pt | Tel. +351 22 8340500 Ext. 1712

Received on Wednesday, 6 November 2002 14:52:27 UTC