Re: Schism in the Semantic Web community.

On 28/1/09 02:12, Alan Ruttenberg wrote:
> On Tue, Jan 27, 2009 at 3:14 PM, Seaborne, Andy<andy.seaborne@hp.com>  wrote:
>> Alan,
>>
>> You quote:
>>> 2.1 Document Conformance
>>> Several syntaxes have been defined for OWL 2 ontology documents, some
>>> or all of which could be used by OWL 2 tools for exchanging documents.
>>> However, conformant OWL 2 tools that take ontology documents as
>>> input(s) must accept ontology documents using the RDF/XML
>>> serialization [OWL 2 Mapping to RDF Graphs], and conformant OWL 2
>>> tools that publish ontology documents must, if possible, be able to
>>> publish them in the RDF/XML serialization if asked to do so (e.g., via
>>> HTTP content negotiation). OWL 2 tools may also accept and/or publish
>>> ontology documents using other serializations, for example the XML
>>> Serialization [OWL 2 XML Syntax].
>> Which says qualifies the "must" in "MUST, if possible, be able to ..."
>>
>> When is it not possible?  (follow-up Q if relevant: who decides?)
>
> The "possible" refers to the cases where RDF/XML can't serialize RDF,
> for instance in the cases that predicates whose URIs localnames begin
> with a number. In those cases one can't construct a qname for the URI
> and since predicates are always written as tags in RDF/XML this just
> can't be serialized in RDF/XML.
>
> So "possible" means possible in the absolute sense.

Another potential work-around here is for the exporting system to 
generate a new (eg. tag: URI) property name, declare it as an 
equivalent-property of the original, and serialize using that. This does 
somewhat distort the original content, of course.

Have any of these problematic properties been spotted "in the wild"? 
Perhaps it would be possible to probe Sindice or other crawlers, 
although of course if the data isn't publish-able in RDF/XML that would 
affect the stats. Still there is N3, Turtle etc data out there.

cheers,

Dan

Received on Wednesday, 28 January 2009 07:54:44 UTC