RE: Comments on Last-Call Working Draft of RDF 1.1 Semantics

On Thursday, December 12, 2013 11:11 AM, Pat Hayes wrote:
> On Dec 12, 2013, at 1:35 AM, Markus Lanthaler wrote:
> >> Two paragraphs later, add a [[sentence]] to the end of the
> >> paragraph:
> >
> > The previous paragraph begins with
> >
> >  "In summary: RDF literals are either language-tagged strings,
> >   or datatyped literals"
> >
> > which is inaccurate IMO. We discussed this before when I wanted to
> > introduce a term for literals that are not langStrings. Here it bites
> > ourselves. Language-tagged strings are datatyped literals
> 
> Weelll not *strictly* they aren't, because *strictly* rdf:langString is
> not a legal datatype. This is why I have to call it out as an exception
> in the semantics and give it its own special semantic condition, sigh.

Hmm... yeah, if you look at it that strictly :-) What a mess for such a
simple feature...

This is completely off-topic and I'm asking it just out of curiosity: What
would break if we would have decided to define a datatype for each language.
So instead of rdf:langString we would have had something like rdf:lang-xxx
similar to the container membership properties rdf:_xx:

  <> rdfs:comment "An explanation in English"^^rdf:lang-en


> > , consequently the OR in this
> > sentence is, strictly speaking, wrong. The simplest way out is
> probably to just remove the whole sentence.
> 
> But I will just omit the "datatyped", and then the contrast is between
> langString and the other cases which combine a datatype IRI with (just)
> one string. OK?

Not sure I follow. Do you want to change that sentence to

  "In summary: RDF literals are either language-tagged strings,
   or literals"

That doesn't make much sense to me.


> >> the datatype it refers to must be specified unambiguously, and must
> >> be fixed during all RDF transformations or manipulations. [[In
> >> practice, this can be achieved by the IRI linking to an external 
> >> of specification the datatype which describes both the components 
> >> of the datatype itself and the fact that IRI identifies the datatype,
> >> thereby fixing a value of the <a>datatype map</a> of this IRI.]]
> >
> > I don't think we need to add this sentence as we provide no mechanism
> > to do so in a machine-processable way anyway.
> 
> True, but it was intended to be another piece of intuitive glue
> attaching "datatype map" firmly to "identifies", which was the real
> point of the changes. Unless you object to its content I would prefer
> to keep it.

I can live with it if you think it's necessary.


--
Markus Lanthaler
@markuslanthaler

Received on Thursday, 12 December 2013 11:16:37 UTC