Re: Revisiting namespaceDocument-8

Norm:  Overall, this seems like a good direction.

That said, one aspect of the namespace issue has been on my mind since we 
"resolved" httpRange-14.  The paraphrase of that resolution is: "an 
information resource can return HTTP 200 and representations; a non-info 
resource should redirect, typically using 303".  So far so good.  If I 
write a poem, I can give it an HTTP URI and serve up representations;  if 
I want to give the poet himself an HTTP URI, then that must redirect using 
303 to a resource that is descriptive of the poet.

Now the questions about namespaces:  is a namespace an information 
resource and therefore the sort of thing that can respond with a 200? 
Certainly it's abstract in a way that the poet is not.  It has no mass and 
won't knock you over if you run into it.   Web Arch characterizes an 
information resource as  one for which  "all of their essential 
characteristics can be conveyed in a message"[1] What are the essential 
characteristics of a namespace, and can they be completely conveyed? 

I note that seem to be abstractions that are not clearly information in 
the Web arch sense.  Consider the concept of the "color red".  I might 
want to make an rdf statement:

        :noah :likes <http://examples.org/colors/red>

Can the essential characteristics of a color be conveyed in a message. 
Does our httpRange-14 resolution allow a 200 status for 
http://examples.org/colors/red?  If not, is a namespace this sort of 
amorphous abstraction, or is it more specifically a collection of 
information, presumably including at least the collection of names 
qualified by a given URI?

If we take the narrow interpretation that the namespace is in fact just 
the set of qualified names, then it's not clear that the representations 
we're proposing to return are in fact representations of the namespace. 
They seem to be representations of a resource that is a description of the 
namespace.  Isn't it exactly that distinction that caused us to argue for 
a 303 in the case of the poet?  As a practical matter, users are going to 
be quite annoyed if we require that for every namespace they deploy a 
server that uses a 303 redirect to get you to the RDDL, but I'm curious 
what our theoretical justification is for allowing a 200 from the 
namespace name. 

Obviously, the last thing I'd advocate is a full reopening of 
httpRange-14, but I do think it's worth convincing ourselves that the 
resolution we've adopted can be applied consistently and conveniently to 
the important case of namespaces and their descriptions.   If we stick 
with our resolution of httpRange-14, then I think we need to be prepared 
to set out what the information is that "comprises a namespace", so that 
we can show that it can be "conveyed", or else we need to suggest use of 
two URI's and a 303 for each namespace.

In any case, I like Norm's proposals for what the representations should 
be and for using GRDDL. 

Noah

[1] http://www.w3.org/TR/webarch/#def-information-resource

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Monday, 27 June 2005 23:24:48 UTC