Re: Q to implementers: Resource identifiers - XML Names and/or(concatenated) URIs? (was RE: rdfs.isDefinedBy...)

On 2002-06-10 15:20, "ext Thomas B. Passin" <tpassin@comcast.net> wrote:

> 
> [Patrick Stickler]
> 
>> On 2002-06-09 23:09, "ext Thomas B. Passin" <tpassin@comcast.net> wrote:
>>> There must be only one way to go from URI references to prefixes and
> back
>>> again. That way should be compatible with XML Namespaces.  It should
> always
>>> be possible to use prefixes as aliases for a "base" URI to make it
> easier to
>>> read and write RDF/XML by hand (I say "base" in quotes to distinguish it
>>> from "xml:base", since the two may not turn out to be the same).  After
> all,
>>> it's done all the time in N3, why not everywhere?
>>> 
>>> Cheers,
>>> 
>>> Tom P
>> 
>> The fact is that qnames and URIs are two competing schemes for global
>> naming, and URIs do not support the structure and contextual semantics
>> defined for qnames -- therefore a fully bidirectional mapping without
>> loss of information is just not possible. URIs will always represent less
>> information than a qname.
>> 
>> The solution IMO is for the RDF/XML serialization should be redone in such
>> a manner that *no* qnames are used to denote resources which are denoted
>> by URIs in the graph. I.e., do away with any need to perform a qname<>URI
>> mapping. Resources are ever and only identified with URIs, whether in
>> RDF/XML or the graph.
>> 
> 
> I think of this as  a purist's approach and although I can sympathize with
> it, and realize that it is the easiest to get unambiguous, I think there's a
> real place for a capability of useful aliases.

These already exist. They're called ENTITYs. E.g.

   ... rdf:resource="&dc;Creator" ...

In fact, the behavior of XML ENTITYs is identitical to that of RDF's
treatment of qnames. The ENTITY/prefix is simply expanded during
parsing, and has no presence in the resulting graph.

> Seems to me that the
> namespaces ought to be able to work for that.

Namespaces, or rather qnames, will *never* work for that because RDF
disposes of the qname structure and thus usage will always be
uni-directional.

Generic round tripping from qname to URI to qname simply cannot work, if
we are to respect the existing standards, which allow Namespaces to be
*any* valid URIRef and which base fragment identifier syntax on the MIME
encoding.

This is not a purist position. This is a realist position. It just
won't work. Period.

> I do agree with you this far - it ought to be possible to avoid using XML
> namespaces and prefixes in RDF/XML syntax if you want to.

Yes, that would be better than nothing, and would also provide
for a sort of imperfect round-tripping where qnames are mapped
to URIs in the graph and output then only as URIs -- though that
might also lead to alot of user confusion since the input/output
will not be syntactically identical (even though it will be
semantically identical).

Cheers,

Patrick 

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Tuesday, 11 June 2002 04:12:11 UTC