Re: Fragment Identifiers and Agent Perspectives

On 10/11/2011 9:14 AM, Jonathan Rees wrote:
> Absolutely right, thanks for pointing this out.

OK, great, you're welcome of course. I'm glad to see we're converging on this.

> However this doesn't substantially change the proposal;

I see it as a substantial change, but I don't think we need agree on 
whether it is.

> it's just a refinement. Just replace "agent" with "syntactic context".

I'm not sure that quite does it, as there seems a bit too much emphasis on 
syntax. In many cases, the intended use of a URI is conveyed by embedding 
it in some language with a "syntax", but it might also be by virtue of 
occurring in a database column, or some other means. Maybe what I'm looking 
for is something like just "context", or "context in which the reference is 
made".

I also think this loses something that I was pushing in my earlier note: 
per the formal mechanisms of RFC 3986, each URI should directly identify 
one thing. That's what I meant by "first class" identification. If I see

 http://example.org/people.html#noah

and ask: "does it identify a part of a document?" I should be able to use 
the retrieved representation, RFC 3986, and the specs to which that 
delegates to get a "yes/no/undefined" answer as to whether that URI 
identifies part of the retrieved document. If the answer is "yes", then 
that URI should not also >directly< identify a person. We may say that it 
indirectly identifies me. It's not a symmetric situation, IMO. If I ask, 
without qualification, what the URI identifies, the answer is "part of a 
document".

Now, it's quite OK per this formulation for the media type registration for 
some type like text/n3 to indicate that the URI identifies a person, in 
which case it does not, in a first class way, identify a portion of the N3 
document. What I object too is implying that a given URI has too referents 
that are first class or direct in this sense.

Noah

Received on Tuesday, 11 October 2011 13:56:36 UTC