Re: ISSUE-58: Scalability of URI Access to Resources

Norm Walsh writes:

> That is, there's nothing about using an http: scheme URI that 
> mandates dereference.

Yes, I think that's right, but it's easily confused with a statement that 
would be wrong.  You can indeed make good use of http-scheme URIs in 
situations where you expect that some common processing can be done 
without dereferencing them.  Derefernce is, as you say, not mandated.

Nonetheless, one of the important characteristics of http-scheme URIs is 
that you >can< try to dereference them.  Furthermore, I think we've been 
clear as the TAG that even though you may know that there is lots of 
processing that can be done using your URIs without doing a dereference, 
it's still usually a good idea to provide representations (I.e. to serve 
representations via HTTP GET) anyway.  Using XML namespaces as an example: 
 the typical XML parser does not do an HTTP retrieval on the URIs of the 
namespaces it encounters, but it's still good [1] to have something like a 
RDDL document out there in case some human user or some other tool wants 
to find out about the namespace. 

I know that's all consistent with what you meant, but it would be possible 
to read your statement in isolation as implying "don't worry about putting 
up representations for your http-scheme URIs.

Noah

[1] http://www.w3.org/2001/tag/doc/nsDocuments/

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

Received on Wednesday, 22 August 2007 19:52:12 UTC