Re: Subgroup to handle semantics of HTTP etc?

Mikael Nilsson wrote:
> ------- Original message -------
> From: Xiaoshu Wang <>
>> I think that semantics should be drawn only from what is asserted in an 
>> RDF content, but we should not draw from how the RDF is obtained.  To 
>> draw conclusion from a network protocol, such as HTTP, essentially bound 
>> URI to its network protocol, which is a very bad idea.
> To me, this amounts to a decoupling of RDF from web architecture, which is, if I understand thing correctly, exactly opposite to the purpose of RDF.
I happens to feel the opposite.  The web has been work so well mostly 
due to the orthogonality between its specification.   In principle, RDF 
is no different from HTML. The content of RDF is created for machine to 
understand just like the content of HTML is created for human (and 
different natural languages for different kind of humans).

But a URI specifies a relationship between a character string to a thing 
in the world, whereas a network protocol specifies how a client and a 
server should *communicate* given a string. How the latter is conducted 
has nothing to do what the former denotes.  Take the ISBN as an 
example.  Should we consider how the ISBN is used to obtain the book it 
denotes be part of the book as well?  Obviously not, right? Then why a 
URI should be any different, if in the future when ISBN is replaced by URI?
> What the server at has to say about the URI is authoritative, according to web architecture, which is not 
> true for *any* other protocol. Therefore any conclusions that can be drawn from that interaction are really interesting.
In semantic web, anyone can talk about anything in anyway.   In other 
words, "authoritative" is a relative concept.  For instance, most people 
use the Dublin Core vocabulary in a different way than what is specified 
(for instance, as a set of annotation properties in protege), is it 
wrong? I don't think so.  But which one is more authoritative? Depends 
on your usage, right?  The web is just like a democratic society, what 
is authoritative is elected by being shared the most but not by what the 
owner says.  But the common practice is if we don't agree with the 
owner's opinion, we just don't use them.
>> URI is just a symbol that denotes a thing in the world.
> I truly hope that this is not true, not even for RDF.
But this is true from the URI spec.
>     The form of URI  is just 
>> such a design of that, given a URI, who we should ask the question about 
>> the URI first without any other knowledge about the URI. 
> Not only ask first, but the final authority about the URI.
Do you mean "final" in a "constant" or "non-modifiable" way?  I hold a 
different opinion on that.  Semantic web is based on an open world 
semantics, hence, no one can say that this is "the description" of a 
resource, nothing more and nothing less.

In fact, what  a URI denotes  is always more than what you get back from 
dereferencing the URI. Dereferencing a URI gets back just *a 
representation* but not *the* representation of what the URI denotes.  
Even if the server never respond with a particular representation of a 
URI, the representation still exist as far as what the URI concerns.


Received on Tuesday, 16 October 2007 15:59:02 UTC