Re: Clarifying Temporal Datasets

On Thu, 2013-05-16 at 10:10 +0100, Richard Cyganiak wrote:
> James,
> 
> On 13 May 2013, at 14:44, James Leigh <james@3roundstones.com> wrote:
> > [1] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html
> > 
> > In section 1.6 Working with Multiple RDF Graphs, I suggest the
> > relationship between datasets and graph be clarified. Particularly wrt
> > temporal changes. Perhaps the following might be better.
> > 
> >        An RDF dataset is a collection of RDF graphs. All but [at most]
> >        one of these graphs have [at least one] associated IRI. They are
> >        called named graphs, and the IRI is called the graph name. [One
> >        of the graphs is called the default graph of the RDF dataset and
> >        might] not have an associated IRI.
> 
> This section is an informative introduction. It attempts to create correct intuitions in a reader, while omitting details. The details are given in the later normative sections, which are clearly hyperlinked from the introduction. What this introduction must not do is simply restate the normative text in different words. I'd rather remove the paragraph than doing that.
> 
> >        [The graph name can be
> >        associated with at most one graph at a time, but that associated
> >        graph may change over time. The dataset's collection of graphs
> >        and associated IRIs may also change over time].
> 
> As Sandro said, that's not correct. Datasets are immutable, like in SPARQL and like graphs.
> 

Fair enough, but I think it should be clarified (somewhere) that the
same graph may have multiple associated IRI and the default graph may
also have one or more associated IRI as well.


> > In section 3.3 Literals, the regex expressions include an '@' prefix,
> > however, the spec never mentions what this '@' is for nor do language
> > tags support an '@' at all. ;)
> 
> Yeah, the note is somewhat misleading anyway as the regex accepts many strings that are not language tags.
> 
> > In section 5.5 The Value Corresponding to a Literal, It says they "MUST
> > accept ill-typed literals". I believe that should be changes to "SHOULD
> > accept ill-typed literals", since earlier it says they SHOULD NOT reject
> > them.
> 
> Sorry, where does it say that they SHOULD NOT reject them?
> 

In section 5.4 Datatype IRIs, "Applications may give a warning message
if they are unable to determine the referent of an IRI used in a typed
literal, but they should not reject such RDF as either a syntactic or
semantic error."

I believe MUST is too strong either way, as many RDBMS database have a
limited set of data types, but one should still be able to use a
domain-specific RDBMS schema to store RDF data.

Regards,
James

Received on Thursday, 16 May 2013 13:15:56 UTC