RE: Presentation Syntax - why?

Thanks Dan.
(my last post was a little out of sync)

I've nearly got it now. The only thing I'm lacking are the descriptions of
what constitutes an exchange syntax, what constitutes an presentation
syntax, the intended usage of each and why it is good that they should be
different...

> I'm pretty ambivalent about this XML presentation syntax; I tend
> to use N3 for that...

Ok, that's a good clue.

> But I've picked up a few bits of motivation for it along the way:
>
>  * you can use XML schema tools to validate documents that use
>   this syntax before you convert them to RDF/XML. Finding typos
>   in RDF/XML documents can be a pain. You write <soc:Preson>
>   and RDF tools just say "oh goodie; a new class I've never
>   seen before. Tell me all about it." but (with a few noteable
>   exceptions such as the DAML validator) they don't say
>   "you probably meant <soc:Person>" or anything like that.

That (IMHO) is a major plus - the inability to properly validate RDF/XML
using regular tools (Relax & co excepted, I believe) is a drawback I'm not
sure is fully compensated for in its flexibility. But in turn I'm not sure
this in the XPS is adequate compensation for the cost of having to deal with
another syntax.

>  * Some folks asked about stream-oriented processing for OWL;
>  with RDF/XML, you pretty much can't get around a need to
>  keep all the triples in memory for the whole parse. With
>  the OWL DL structures, I gather you can throw some bits
>  away as you go.

Interesting. But I would be surprised if it weren't possible to add
syntactical constraints to RDF/XML such that the same were possible (and
without having to modify existing style conventions).

> >  If the
> > >underlying model used by the AS&S is compatible with the RDF
> > graph/triples
> > >model, then why not use RDF/XML?
>
> For interchange, we are using RDF/XML.

Interchange between what kinds of agents? XPS SHOULD, MUST or COULDN'T
CONCEIVABLY be used for interchange between machines?

> For folks that like to see the XML syntax more closely match
> the concepts in the ontology, this is an alternative that
> can be mapped to RDF/XML.

Ok. It sounds better put like that. I can see that the OWL ontology being on
top of RDF could have a nature that was so different that it made this
worthwhile. I'm not entirely convinced that there isn't some
quart-in-a-pint-pot jamming going on with RDF+XML+FOL+DL+Graphs+DAML++, but
I guess such situations are a fact of life. Or vice versa.

> >  Or is there such an air gap between the RDF
> > >and OWL layers, that the OWL can fly free with it's own model, syntax
> > and
> > >serialization?
>
> I suppose time will tell.

Thank you.

> > >On a practical level, the question is simple if a developer wishes to
> > build
> > >a Semantic Web application, where do they start?
>
> I hope they start with the interesting bits of their application.
>
> If they want to interchange, they better support RDF/XML.
>
> >  If they start with RDF now,
> > >will they need a rework to be able to include OWL features without
> > the
> > >struggle of [3]?
>
> I don't think so. I think using horn rules with RDF triples is
> pretty straightforward.

(I can't wait to see the Prolog front end ;-)

> >  Or if they start with OWL AS&S will they lose the
> > >compatibility with existing RDF data without building tools to carry
> > out the
> > >transformations of [2]?
>
> Yes, I think so. But other folks think otherwise, and their
> stuff seems to work too.

I'm not sure how justified it will prove to be, but I do feel reassured by
your tinkerer's optimism ;-)

Cheers,
Danny.

Received on Tuesday, 17 June 2003 20:39:14 UTC