W3C home > Mailing lists > Public > semantic-web@w3.org > August 2008

Re: namespaces

From: Richard Newman <rnewman@twinql.com>
Date: Fri, 29 Aug 2008 15:44:51 -0700
To: "Richard H. McCullough" <rhm@pioneerca.com>
Message-Id: <25810A38-FECF-47E4-AE11-1890F05C24CD@twinql.com>
Cc: "Semantic Web at W3C" <semantic-web@w3.org>

> According to section 2.1 of the OWL Web Ontology Language Guide,
> an ontology begins with namespace declarations.  For example
>   xmlns:rdf ="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
>
> I identify "rdf" as the user-friendly name of the context.
> I identify "http://www.w3.org/1999/02/22-rdf-syntax-ns#" as the URI  
> of the context.
> I identify the propositions contained in the RDF file returned by  
> this URI
> as the propositions which are the context.

Namespace declarations like that exist to abbreviate (and, in some  
cases, make possible) RDF/XML output. They allow you to encode URIs as  
QNames.

Even trivial processing of a chunk of RDF/XML can result in those  
namespace declarations being stripped away or renamed. Other  
serializations, such as N-Triples, don't even have namespace  
declarations.

The namespaces you are discussing are purely syntactic artifacts.  
(Linked Data advocates dereferencing the URI of a resource itself, not  
some notional 'namespace'.)

Consider that:

  (a) not all URIs are dereferenceable. I could have
     xmlns:foo="urn:baz" in the document.

  (b) dereferencing a URI can result in a variety of redirects. A  
fetch for that RDF URL could return a 404, or a 302 or 303 to a  
different URI. I have no idea what that means in your scheme. How  
would you handle redirects and Content-Location headers?

  (c) the representation retrieved might not be RDF. It is entirely  
reasonable (in the context of the Web) for the only available  
representation of the RDF specification to be a PDF, or HTML.

  (d) in the case of an RDF representation being received, the graph  
is not necessarily static.

All of these are true for resource URIs in general, but especially so  
for namespace abbreviations.

This means you might end up with 'contexts' which cannot be retrieved,  
have multiple identifiers and friendly names if they can be, and vary  
over time.

What do you hope to gain by your assertion? What response are you  
looking to receive from this community?

What is your formal definition of a context in this assertion?

How is this of interest to the folks on the Semantic Web list?

-R
Received on Friday, 29 August 2008 22:45:28 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:24 GMT