Pierre-Antoine CHAMPIN wrote:
> 
> I think this raises a question about the semantic relation between Literal and Resources.
> 
> As I understand RDF, there is a dichotomy between both,
> 
> Some people on this thread talked about 'dereferencing',
> but I don't quite agree with teir view :
> I don't think we NEED to dereference, for example, an RDF file's URI to its RDF content :
> RDF statements are about resources, and the resource IS the content, am I wrong ?

The semantic of the URI is not predefined. It could represent a thing,
not connected to the web. It could represent a person. In those cases,
the content of the URI is NOT the resource.

But it would be nice if there was a standard as to the
intrepretation. In this way:

  * If the URI is recognized as an retrievable document, the resource
    IS that document.

  * If the URI is not an retrievable document (an URL), it's abstract
    for you.

There is no way the core RDF could know if an URI is a document that
could be accessed for the extraction of more data. The implementation
has to analyze the URI. If URI schema turns out to be http, you know
you can (try to) retrieve the document. But you don't know the type of
document until you get a mime type.

There will be many schemas other than http. In the future, maby
there are a global isbn network that actualy let you access the
books.  Maby there will be a social-security-number schema.

Any given implementation will only support a subset of all the types
of URIs. It could know what types it can retrieve, what types
represent "the outside world", etc. But new types could be
invented. Previously unaccessible objects, like bar-code numbers,
could be made accessible by a new internet service, giving metadata
about the product.

There could easily be portals for metadata about this or that
thing. Places, ideas, or anything else.

The connection of the URI to "more information" is a convention. For
http, there is the system of using domains, etc to access the
data.  But new conventions could be contructed, based on other
models.


The conclusion:

A way of handling the 'dereferencing' of resources would have to build
on an RDF model of the URI itself.