3 Reasons for not having XMLLiteral in DL

Hi -

sleeping over our discussion, I found several
reasons, why we should not have XMLLiteral in
DL.

1)	Usability aspect: Increase of modelling alternatives

Users are given a whole new means to model information,
whose semantics unfortunally will be outside of the logic,
e.g.

(Raphael, type, Person)
(Raphael, firstname, "Raphael")
(Raphael, hasAddress, [XMLLiteral]
	<addresses>
		<address>Sidbury Rd, Chorlton cum Hardy, UK</address>
		<address>Kolberger Str, Karlsruhe, Germany</address>
		<address>Kirchhaldenweg, Loffenau, Germany</address>
	</addresses>
)

vs.
(Raphael, type, Person)
(Raphael, firstname, "Raphael")
(Raphael, hasAddress, addr1)
(Raphael, hasAddress, addr2)
(RAphael, hasAddress, addr3)
(addr1, street, "Sidbury Rd")
(addr2, street, "Kolberger Str")
(addr3, street, "Kirchhaldenweg")
...

What are the design rationales/trade-offs for modeling
something as a XML fragment vs. as part of the OWL KB ?

2)	Nesting RDF

What does it mean, if a new RDF modell is nested within
a XMLLiteral, are the triples inside those models part
of the containing RDF Model or do they constitute a
new models, hence a RDF graph is not a labeled pseudograph
but a much more expressive structure (cannot find the
corresponding English translation), that allows nesting graphs

(Raphael, type, Person)
(Raphael, firstname, "Raphael")
(Raphael, authorOfOntology, [XMLLiteral]
	<rdf:RDF>
		(x, type, Ontology)
		(x, imports, foo)
		(y, type, Class)
		(z, type, ObjectProperty)
	</rdf:RDF>
)

does this translate to
(Raphael, type, Person)
(Raphael, firstname, "Raphael")
(Raphael, authorOfOntology, [XMLLiteral] ...)
		(x, type, Ontology)
		(x, imports, foo)
		(y, type, Class)
		(z, type, ObjectProperty)

or

(Raphael, type, Person)
(Raphael, firstname, "Raphael")
(Raphael, authorOfOntology,

	(x, type, Ontology)
		(x, imports, foo)
		(y, type, Class)
		(z, type, ObjectProperty)
)

4)	Argument of opening OWL to multitude of other XML technologies
	is "broken" (to speak in PPS terminology)

The only thing we do is enabling the embedding of arbitrary XML,
this again is susceptible for actually decreasing interoperability,
since vendors will use this as means for distinguishing each other.
For example, some vendors will certainly make use of schema validation
of XML fragments in XMLLiteral (and mandate this in their tool), hence
semantic aspects will be outside of the processing of other tools.

Of course your counter argument will again be (as ever so often  before)
that we cannot prevend this from happening in the first place, however
I strongly believe that we should try to facilitate other behaviour where-
ever we can.


Mit freundlichen Grüßen,
Best regards,

Raphael Volz
Institut AIFB, Universität Karlsruhe
http://www.aifb.uni-karlsruhe.de/WBS/rvo
volz@aifb.uni-karlsruhe.de

WIM, FZI Karlsruhe
http://wim.fzi.de/
volz@fzi.de

Fax: 01212-5-470-17-365

Received on Wednesday, 5 March 2003 10:00:36 UTC