Re: [xml-dev] What is RDF for? (Was: What are Web Services for?)

Matthew Gertner wrote:

> Murray Spork wrote:
[...]


>>[1] http://archive.dstc.edu.au/RDU/staff/jane-hunter/www10/paper.html
>>"Combining RDF and XML Schemas to Enhance Interoperability Between 
>>Metadata Application Profiles"
>>
> 
> Cool, they seem to be proposing almost exactly the same approach as I am.
> Thanks for posting this. The only significant difference is that I would
> prefer to see the RDF-type semantics (i.e. the ones that can't be expressed
> using document schemas) in a separate annotation file, rather than embedded
> in the schema. First of all, this divorces the information from the specific
> (XML) schema syntax being used. Also, it is a cleaner and more modular
> approach.


Agreed - something like the SAF 
http://www.tibco.com/products/extensibility/saf/exec-summary/

 
> In general, I think that it makes a lot of sense to modularize the semantics
> expressed in a document schema. In fact, XSD would have benefitted greatly
> from this, since they could have created a core language for describing
> document structure (which would have ended up looking a lot like RELAX NG,
> at least in expressive power) and offloading all of the other gunk into
> annotations that also use a standard syntax. As far as using little snippets
> of RDF syntax inside the annotations, as the paper proposes... why not?


> How come you don't see significant benefits in this? 


Sorry - poor use of english on my part - my current thinking is that 
there are indeed significant benefits to be obtained by this approach.

> Why is it better to
> create two separate schemas with significant overlap, and rob RDF-style
> query engines of the semantics already in the document schema (or do hybrid
> RDF/XSD parsers how exist?).

I'd like to know this as well.

As you and the DSTC article have pointed out, there is overlap between 
RDF and Schema - so how do you demarcate between the 2? The dstc article 
proposes a set of guidelines to follow so that overlap or duplication of 
semantic information does not occur (which whilst being a more immediate 
solution; I don't feel it is ideal). The other approach is the one you 
suggest here - a hybrid parser. The DSTC article raised this possibility 
as well. You'll still get conflicts - but maybe they can be managed?

Regards,

--
Murray Spork
CITI
The Redcone Project
Queensland University of Technology, Brisbane, Australia
Phone: +61-7-3864-4246
Email: