Re: Uniform access to descriptions

Williams, Stuart (HP Labs, Bristol) wrote:
> Hello Alan,
>  
> I really wish that we could 'nail' this kind of question...
>  
> I happen to prefer to think of things in this way...
>  
>     That the questions asked of the web in the form of, say, HTTP GET, 
> HEAD... operations are questions asked of the web and that it is the 
> web that responds.
>  
>     The web, broadly offers 4 kinds of answers:
>  
>         - Here's a awww:representation of what you asked about (2xx 
> responses)
>         - Asking a suggested different question may yield a better 
> answer (3xx responses)
>         - SOL (4xx responses)
>         - Something in the infrastructure is broken (5xx responses)
>  
> This completely avoids having to attribute responsive behaviour to the 
> things of the web themselves, though the web is able in some sense to 
> 'inspect' at least some 'things' inorder to formulate its 
> response. Instead one attributes responses to the infrastructure of 
> the web itself which, afterall, is what creates the illusion that a 
> sequence of magnetic domains organised in a sectored radial fashon on 
> a disk is a document (any document let alone a Microsoft Word 
> document). It also avoids other quagmires... which I'll avoid now my 
> not mentioning further :-).
That would works fine.
>  
> Just a question about this document that you have in mind:
>     How are you identifing it? [eg. by name, say in a filing system; 
> by bit/sequence of its serialisation...]
Let me suggest to distinguish the meaning of "denote" and "identify".  
Hence, a URI is a name that *denotes* a resource.  But a *resource* (not 
URI) can be identified in many ways.  For instance, we can *logically* 
identify a resource by a set of logic assertions, or casually 
*identified* in natural languages, pictures, etc. 

A URI can be bound with any transportation protocol to retrieve various 
representation that can be used to *identify* a resource.  Within the 
web, we can then say that a URI denotes a "resource" but a particular 
HTTP GET of a URI *identifies* the resource's representation, which is 
obviously in its electronic form.  But URI doesn't have to be bound with 
HTTP.  Our snail mail system might in the future use URI too.  But by 
snail mail GET of a URI would *identify* a different *representation* of 
the resource denoted by that URI.  But this time, it could be in print 
or in some other medium form.

I hope this might help.

Xiaoshu

Received on Wednesday, 9 April 2008 12:46:12 UTC