Re: RDF/XML and CURIEs

RDF/XML is the only format that a SPARQL implementation must understand. 
Whether a particular SPARQL implementation understands Turtle (the obvious 
alternative) is another matter. Crazy ivan uses sparqler and virtuoso; I 
would not be surprised if both did.

Having said that: I do not really think that the problem is as huge for us 
now as you seem to think. These issues are relevant for predicates only; 
the amazon example you give probably not lead to a predicate but to a 
complex URI for a subject or an object, where the problem does not occur...

I.


----
Ivan Herman
W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
Foaf: http://www.ivan-herman.net/foaf


_____ Original message _____
Subject:	Re: RDF/XML and CURIEs
Author:	"Shane McCarron" <shane@aptest.com>
Date:		10th March 2008 5:10:51 

Then I guess I need to ask a different question.  Given that there is 

this truly awful issue with RDF/XML, is there some other format that 

SPARQL will swallow?  "cause I really do not want to go to the trouble 

of making my parser generate an output format that in some large number 

of cases just wont work.  That would be crazy.

Ivan Herman wrote:
> Hi Shane,
>
> congratulations! You have discovered one of the nasty little secrets 
> about RDF/XML, namely that there are perfectly valid RDF graphs that 

> cannot be encoded into RDF/XML. This is a known problem, filed as an 

> official unsolved issue for RDF/XML that a future RDF group (if such 

> an animal will be created) will have to take care of.
>
> Note that the same RDF graph can be expressed in NTriples and in 
> Turtle without any problem. And, well, in some ways, in RDFa (except 

> that the RDFa parser should output Turtle, for example, to be 
o.k.).
>
> So: if by 'we' you mean the RDFa group, then the answer is indeed that 
> we do not care. If by 'we' you mean the Semantic Web activity than, 

> well, yes we have to care about serializations in general...
>
> Ivan
>
> B.t.w.: this is the very problem Misha Wolf and IPTC had for a long 

> time...
>
> Shane McCarron wrote:
>>
>> Okay.... Someone please tell me I am wrong about the 
following:
>>
>> When trying to force my RDFa parser to emit RDF/XML so it will work 

>> with the test suite, I noticed that the general model for RDF/XML is 

>> that predicates are expressed as elements.  E.g. a property of 
>> "foaf:name" in RDFa would map to an element <foaf:name> in RDF/XML.  

>> And somewhere the foaf namespace would be declared.  So far so 
good.
>>
>> Then I started thinking about our extension model and CURIEs.   If I 

>> use one of our existing examples about isbn numbers..... let's say I 
>> declare xmlns:amazon="http://www.amazon.com/books?isbn="  and in my 

>> RDFa I have something like <span property="amazon:12345">My Book 
>> Title</span>
>>
>> I think that we generate a triple where the predicate is 
>> amazon:12345, or http://www.amazon.com/books?isbn=12345 - again, so 

>> far so good.
>>
>> But, if I want to map that into RDF/XML, I need to do something 
like:
>>
>>    <amazon:12345>My Book Title</amazon:12345>
>>
>> Sadly, that is invalid.
>> Please, someone tell me I am wrong.  Alternately, someone tell me we 

>> don't care.
>>
>

-- 
Shane P. McCarron                          Phone: +1 763 786-8160 x120
Managing Director                            Fax: +1 763 786-8180
ApTest Minnesota                            Inet: shane@aptest.com

Received on Monday, 10 March 2008 16:53:16 UTC