Re: advice on modeling an XBRL fact as a triple

Hi Bob,

> How does the following look?
>
>   _bnodex rdf:type "us-gaap:SharesOutstanding";
>           xbrl:contextRef "x1";
>           xbrl:unitRef "Shares";
>           xbrl:decimals "-6";
>           us-gaap:SharesOutstanding "485000000".
>
> If the use of the rdf:type object as a predicate is not a good 
> idea, is there some convention (i.e. an existing predicate from
> an existing ontology) to represent the PCDATA of an element that
> has attributes that are also being represented by triples?

If the concepts in XBRL taxonomies are classes, then rdf:type would 
seem to be appropriate, see:

   http://www.w3.org/TR/rdf-schema/#ch_type

In general using the RDF schema predicates for type, subClassOf, 
subPropertyOf and label would be useful for modeling XBRL 
taxonomies.

One question is the namespace for describing XBRL instances. Is it 
better to the preexisting xbrli <http://www.xbrl.org/2003/instance> 
namespace instead of a generic XBRL namespace?

It might be worth adding datatypes to the literals, e.g.

         "-6"^^xsd:integer

A further question is how to make the URIs for RDF resources 
identified from XBRL independent on the website on which a copy of 
an XBRL filing resides.

XBRL Linkbases use XLink to refer to element definitions in XBRL 
Schemas. The URI for the schema depends on the server it has been 
copied to, e.g. the EDGAR archive. It may be better to use the 
target namespace for the schema and element name, in place of the 
schema URI and element definition id.

This begs the question of what is the namespace for relationships 
declared in XBRL linkbases.

  Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett

Received on Monday, 15 September 2008 13:30:49 UTC