Re: resources and URIs

Well, I think we're making progress here.

pat hayes wrote:

>> The Web Architecture doesn't provide a way to talk about what 
>> resources are or what they mean.
> 
> Why does it even need to mention resources at all?
...
>> This is all very well and good, but it is neither necessary nor 
>> appropriate that we pretend that the current architecture of the web 
>> comprises any notion of "about-ness".
> 
> Fine: but then please don't talk about semantics in the architectural 
> description. Why do you even need to mention resources which the 
> representations are 'of' ?? All the architecture needs to do is to ship 
> representations around. Leave what they are 'of' to a semantic theory.

 From the point of view of Web Architecture as discussed in the current 
draft, a resource is simply "that which is identified by a URI".  This 
is highly consistent with the 2396 language "anything that can be named 
or described".  I accept that this notion has almost no semantic weight 
and is probably not essential to the argument, but I don't think it's 
appropriate at this point for our document to stop using the word 
"resource", which has been so central to the web for so long.

Maybe I'm missing something, but it seems to me that there ought to be a 
smooth continuum between the Web of today, which says exactly zero about 
resources other than they are identified by URIs, and the Semantic Web, 
which addresses notions such as about-ness and denotation.

This leaves one severe friction point; let me try to expound it to see 
if we are at least clear on the area of disagreement: webarch currently 
says that URIs may be used to retrieve representations of resources. 
Reading this according to the normal meaning of English words causes me 
no stress: if you claim that an HTML page is a representation of the 
weather in Oaxaca, I feel no discomfort.  You, on the other hand, are 
profoundly disturbed by the notions that (a) the weather in Oaxaca can 
be a resource, i.e. an object in a networked information system, and (b) 
that an HTML page can be a representation of it.

Do I have this straight?

> There is no way to know if it is correct or not since it doesnt seem to 
> mean anything that can be tested for truth.  Am I identified by a URI? 
> How could anyone possibly tell? 

You're right; the current web architecture provides no way to test this 
condition.  I had (perhaps naively) understood that the Semantic Web was 
going to give us the machinery to make machine-usable assertions about 
semantic classes of resources and their relationships.

> I don't think it does, in fact. That is, I think that it is necessary to 
> make this distinction even to make proper sense of examples like the 
> weather in Oaxala. If this distinction were not made, websites like 
> http://www.wunderground.com/US/FL/Pensacola.html would be incompatible 
> with the current TAG account of Web architecture, or else weather would 
> have to be part of the information network.

Here we part company.  That website makes perfect sense to me, and if 
its owner wants to explain (in English) those pages as representing the 
weather in Pensacola, that causes Web software no heartburn.

> If this is true, then please 
> tell me what transfer protocol  and MIME type I should use for accessing 
> the weather on the Web: not information *about* the weather - as you 
> have made clear, anything to do with aboutness is not part of the 
> architectural description -  but the weather *itself*.

You can't access anything on the web.  You can only dereference the 
identifiers to get representations of that which they identify.

> Fine, but what IS that sentence saying? What do y'all mean by "a 
> representation of" a resource? In both the informal English senses and 
> any technical sense I know of, that has to mean that the representation 
> *represents* the resource. Apparently you do not mean that. So what DO 
> you mean?

The architecture defines a formalism.  In the formalism, there are 
resources, about which we say effectively nothing, identified by URIs on 
which we impose syntactic constraints.  Rules are provided for some URI 
schemes whereby protocols are used to send and receive representations, 
formally specified as packages of electronic information comprising 
metadata and payload.

It is not that uncommon, in computer science, for formalisms to be given 
names suggestive of their application-level usage; examples would 
include "object", "method", "header", "interrupt", and so on.  From this 
point of view, the formalisms of "URI", "resource", and "representation" 
are empirically observed to be sufficiently well-specified as to serve 
as the basis for the successful construction of software.  The English 
names for the formalisms strike me as sound, rhetorically and tutorially.

I think you are suffering entirely unnecessary angst from overloading 
excessive semantic weight on what at the end of the day are just labels 
for the parts of a successfully functioning formalism.
-- 
Cheers, Tim Bray
         (ongoing fragmented essay: http://www.tbray.org/ongoing/)

Received on Wednesday, 16 July 2003 22:37:40 UTC