Re: dcterms:creator in RDFa Core 1.1, WD of October 26

On Wed, Nov 10, 2010 at 10:11:22PM +0000, Toby Inkster wrote:
> > Some of the examples in [1] generate triples such as:
> > 
> >     <> foaf:primaryTopic <#bbq> .
> >     <> dcterms:creator "Jo" .
> > 
> > However, http://purl.org/dc/terms/creator has a range of
> > dcterms:Agent. Using dc:creator would not be incorrect because
> > http://purl.org/dc/elements/1.1/creator has no specified range
> > (or rather, by default, rdfs:Resource).
> 
> This seems to me to be a conflict between:
> 
> 1. the desire to use the latest and greatest version of Dublin Core; and
> 2. the desire to use a literal object so that the example includes both
> a literal and a URI.
> 
> Perhaps something like dcterms:modified could be used in the example
> instead of dcterms:creator, as that has a literal object?

If an example using literal object is needed at that point in
the spec, that would of course work.  However, that does not
address the question of best practice regarding the fifteen
core properties in their /terms/ and /elements/1.1/ variants.

DCMI assigned ranges in order to conform to emerging notions of
best practice.  However, it did not "deprecate" the rangeless
properties because they had at that point already been used
very extensively in Semantic Web data -- with both literal
and non-literal objects.  By coining "parallel" properties,
DCMI offered data providers a choice.

DCMI has been "gently promoting" the /terms/ variants for being
"more precise" and for helping data consumers by making the
data more consistent.  However, I have been getting feedback
that "rangeless" properties are actually useful, which
Jeni Tennison's point seems to reinforce.

Specifically with regard to Dublin Core properties, it is clear
that the distinction between dc:creator and dcterms:creator is
not widely understood.  If the dcterms: variants are promoted
(e.g., in the RDFa specs), what is the risk that they will
be used incorrectly?  If the risk is high, should we not then
promote the dc: properties?

More generally, what is the message for vocabulary developers
today?  Are there uses for which the right design decision is
to err on the side of underspecification?  This is a general
question of best practice for Semantic Web, but Jeni is raising
it specifically for the case of RDFa.  If not on this list,
where might we have this more general discussion?

Tom

-- 
Tom Baker <tbaker@tbaker.de>

Received on Thursday, 11 November 2010 00:51:14 UTC