Re: PROPOSAL to close ISSUE-39: RDFa term mapping triples

Hi Toby,

On 5 Oct 2010, at 20:47, Toby Inkster wrote:
> Some of my concerns are outlined in
> <http://www.w3.org/mid/bd11e07b6857681b4c1d25e21df964b0.squirrel@goddamn.co.uk 
> >.

Thanks. I will speak to those as best as I can. If you raised  
additional issues in the conference call, and you want me to address  
those, then I ask you to kindly outline them in another email.

So in my understanding, your concern is that given a term mapping

   ex:foo rdfa:term "foo" .

there could be a triple stating

   ex:foo owl:sameAs ex:bar .

and hence one could infer

   ex:bar rdfa:term "foo" .

and now the expansion of the "foo" term is no longer well-defined.


I can understand this concern from the point of view of simply  
considering what individual triples say. But triples aren't free- 
floating in semantic space. They always appear in a context. You  
cannot usefully think about RDF without taking that into account.

First of all, check the proposal [1]. It makes very clear that only  
triples in the profile graph are considered. Hence, the possible  
existence of owl:sameAs triples anywhere except in the profile  
document cannot be an issue.

Second, the primary purpose of term mappings is to establish short  
names for classes and properties. The OWL spec does not recommend the  
use of owl:sameAs for classes and properties, but instead has  
dedicated properties owl:equivalentClass and owl:equivalentProperty,  
which do not exhibit the problematic inference.

Third, what is the worst thing that could happen in the unlikely case  
that a profile author uses owl:sameAs regardless? Note that the  
proposal does not prescribe the use of the OWL entailment regime, so  
the spec does not indicate that any triples should be inferred at all,  
and processing would not be altered.

Fourth, but what if for some unlikely reason an implementor insists on  
applying OWL entailment to the profile graph containing the work of an  
owl:sameAs-abusing profile author? An additional triple would be  
inferred, causing a clash in the term mapping. The proposal states  
that in such a case the term mapping would be ignored, hence in the  
document that used the @profile, the term "foo" would be unmapped, and  
some triples that were perhaps wanted by the document author would not  
be generated. This is the pathologic worst-case scenario, and AFAICT  
the only situation where there *is* actually a problem.

> Even if those concerns didn't apply to terms at all, I still think
> there is value in keeping term mappings as-is for consistency
> with prefix mappings.

I have not weighed in on the question of expressing prefix mappings in  
RDFa, because I don't consider that question critical. I would be  
happy with either option -- keep them as is, or change them along with  
the term mappings. Whatever finds consensus.

Best,
Richard

[1] http://www.w3.org/2010/02/rdfa/wiki/ProfileSpec


>
> -- 
> Toby A Inkster
> <mailto:mail@tobyinkster.co.uk>
> <http://tobyinkster.co.uk>
>

-- 
Linked Data Technologist • Linked Data Research Centre
Digital Enterprise Research Institute (DERI), NUI Galway, Ireland
http://linkeddata.deri.ie/
skype:richard.cyganiak
tel:+353-91-49-5711

Received on Wednesday, 6 October 2010 02:50:38 UTC