Do non-prefixed values of @property generate triples?

SWD Issue Tracker wrote:
> ISSUE-61: namespace for non-prefixed values of @rel
> 
> We've been dancing around the following for a while:
> PROPOSAL - "RDFa parsers should ignore non-prefixed values of @rel that
> are not part o the reserved words."
> 
> See telecon discussion record of 2007-10-04
> http://www.w3.org/2007/10/04-rdfa-minutes.html

This issue raised quite a bit of discussion on the telecon this morning
and I wanted to make sure to follow it up with what seemed to be a
resolution and the deeper discussion that it created about @property and
generating triples.

I think we all agreed that non-prefixed values in @rel and @rev should
not place triples into the default graph. The only exception to this
rule is that values defined in http://www.w3.org/1999/xhtml/vocab#, such
as 'prev' and 'next', do generate triples and place them into the
default graph.

Discussion then followed to how explicit we should be when stating this
in the Syntax document. There seemed to be a general consensus that the
language was already in the document:

http://www.w3.org/MarkUp/2007/ED-rdfa-syntax-20070921/#processorconf

However, Mark was concerned that this was slightly different and had
more to do with CURIEs than processor conformance, and needed to be
stated explicitly, rather than implicitly. Ben was concerned that we
would be repeating ourselves if we put it into the document. Mark was
asked to draft some language to see where and how it would fit in.

This raised another issue, however, which was:

Do we apply the same triple generation rules to @property that we have
defined for @rel and @rev (no prefix = no triples generated), or do we
special-case @rel and @rev and any value in @property generates a
triple. For example, should this generate a triple if 'foo' is not defined:

<div about="#muse-knights-of-cydonia" instanceof="hmedia:Recording">
   <span property="dc:title">Knights of Cydonia</span>
   <span property="singer">Matthew Bellamy</span>
</div>

We know that the following triples are generated:

<#muse-knights-of-cydonia>
    rdf:type hmedia:Recording ;
    dc:title "Knights of Cydonia" ;

The question is, should the following triple also be generated and
placed in the default graph:

<#muse-knights-of-cydonia>
    singer "Matthew Bellamy";

I would argue that the "singer" triple does not need to be generated for
an RDFa conformant parser for the following reasons:

 - In the interest of RDFa adoption, the rules that we create should be
   orthogonal. The same CURIE processing rules that apply to @rel and
   @rev should be used for @property. Every special case we create will
   cause more confusion for publishers and RDFa adopters.
 - From a philosophical standpoint, singer actually doesn't mean
   anything because it is undefined. I don't think that the basic RDFa
   parsing rules should generate triples for things that do not have a
   defined meaning.
 - There is nothing preventing a RDFa conformant parser from generating
   the triple anyway, but according to the RDFa Syntax document
   Conformance section, it must place it in a graph that is not the
   default graph.

-- manu

-- 
Manu Sporny
President/CEO - Digital Bazaar, Inc.
blog: Bitmunk Launches World's First Open Music Recommendation Service
http://blog.digitalbazaar.com/2007/09/09/bitmunk-music-recommendation/

Received on Thursday, 4 October 2007 17:47:22 UTC