>From a mail to the list last week: > > Question: is the definition of ParameterInstance wrt data types > ill-defined? > > ParameterInstance resource definition table contains this in the > description of rdf:value: > "The value of the parameter. rdf:datatype SHOULD be used to indicate the > type of the parameter instance value." > > Yet rdf:datatype is only defined in the context of RDF*/XML*, where it > gives the data type of a typed literal - and that in turn would affect > comparison results in other RDF specs like SPARQL, so this is not an > empty/nop question/series of them. >
The intent here was consistency with the RDF Concepts definition of Datatypes [1] which describes it as an abstraction compatible with the definition of datatypes in XML Schema Part 2. I agree that the RDF Syntax definition of rdf:datatype [2] does constrain it to RDF/XML. Possible alternate wording: "If the value is a literal, the datatype for the literal should be indicated using the syntax appropriate for the serialization, if one exists.". Other suggestions? > ParameterInstance's resource definition table row for rdf:value leaves the > value unconstrained; in particular, the value (object of the rdf:value > triple) need not be a literal at all. If it is not a literal, e.g. it is > a resource reference, it is incoherent to talk about the type of said > (not-) literal. The intent here was consistency with the RDF Schema definition of rdf:value [3], but as I re-read it I think the Automation spec is incorrect. RDF Schema defines both the domain and range of rdf:value as rdfs:Resource. > > Clearly we should not be using RDF/XML-specific syntax to put requirements > on the RDF. My best guess at the intent from what is written would be: > - If the value type is a RDF literal, then it Should be a RDF typed > literal. [that buys you rdf:datatype when the media type is > application/rdf+xml] > - If the value is a resource, ??? I'm guessing no added constraints ???. Your best guess at the intent is correct. If it is a literal, help the consumer understand how to interpret it. If it is a resource, the consumer will likely have to inspect the resource to understand it (or perhaps inspect its resource shape). Suggestions on how we can make this less ambiguous in V2? [1] - http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/#section-Datatypes [2] - http://www.w3.org/TR/REC-rdf-syntax/#section-Syntax-datatyped-literals [3] - http://www.w3.org/TR/rdf-schema/#ch_value
