Lists Home |
Date Index |
> ... but how would we know when it does and doesn't return a
> pointer to something physical? ISBN might return a URL, but it also
> represents non-url things - i.e. a physical book.
Most of the time the context will make clear whether a URI is there for
accessing a netwrok-accessible resource or not. I think this notion has
been getting lost in the discussion.
For example, say a namespace identifier is in the form of a urn. The
processor may try to use it to retrieve a schema, but an xml schema
processor knows that may fail (being only a hint). Otherwise, processors
know the namespace identifier is essentially a string, so there is no
problem about deciding to sit or not. The same applies even if the
namespace identifier is in the form of an http: url.
Say a particular identifier is used to specify the type of a resource in
RDF. The RDF processor does not need to access that url to work with it.
The processor may need to know where to look up the definition of some verb
(predicate), but that is a different thing altogether. Suppose there were
an RDF triple that declares the home page of a resource:
Here, only the object, the literal value of the url for the homepage, is a
candidate for accessing. This is not dependent on human judgement, it would
also be known to the processor, regardless of whether the other two URIs
used the http: scheme or not.