Re: TB16 Re: Comments on arch doc draft

On 2002-07-03 20:27, "ext Jonathan Borden" <jonathan@openhealth.org> wrote:

> ... and I believe that a
> real web architectural principle is being discussed (actually I consider
> this the single most important web architectural principle)

I wholeheartedly agree.

> The archetectural principle of the Web is that, roughly, the spelling, or
> what characters that are used to represent the word, e.g. "bad" or
> "butterfly" are not important because we can obtain a _definition_ of the
> word by _dereferencing_ it.

Ummm... sorry, but I missed where that was defined. Precisely where
does it say that one should be able to get the definition of a URI
from the URI itself?

Where exactly is that defined as the "architectural principle of the web"?

>> 
>> Most people (including me) think that http: scheme URIs should be used
>> for WEB PAGES.  Even Paul Prescod would agree that http: scheme URIs
>> should be used exclusively for resources which are interacted with
>> through the standard HTTP verbs.
> 
> What do you mean "URIs should be used for WEB PAGES"? Do you mean that when
> you dereference a URI you should get a Web page back? The point is that a
> URI can easily be used to represent a "car" just as the word "car" can be
> used to represent a car. Of course neither are _actually_ a car. And when
> you type the characters {'c', 'a', 'r'} into an online dictionary, or type
> the word, just for you Joshua :-)) into MS Encarta, you might get back a
> piece of text that describes the word "car", and/or a picture of a "car".
> This is a description or representation of the word "car", but certainly
> neither _is actually_ a car ... if this is so obvious, what is the
> difficulty in having _any_ URI refer to the concept of a "car", and yield a
> description or representation when dereferenced?

Because if the same URI denotes both the car and a picture of a car,
you cannot make statements about either without ambiguity, and therefore,
though that might be sufficient for the wild-and-wooly human-focused
Web, it's anathema to the Semantic Web, which requires a far higher
degree of precision.

This debate, as I see it, really boils down to a conflict between the
Web and the Semantic Web.

The Web is all about getting access to things. It doesn't much care
about stuff that can't be accessed online. If the thing itself can't
be accessed, such as a car, humans are happy enough to get a picture
or description of the car, and are usually savvy enough to understand
the difference and handle the ambiguity.

The Semantic Web is all about saying anything about anything, and
doesn't care anywhere near as much as the Web about whether those
things are accessible online or not. What it *does* care about is
that if you say something about something, it is clear and consistent
what you are talking about.

The attached set of N3 (RDF) statements illustrates this critical
need of the Semantic Web for avoiding any overloading of URI
denotations.

> To be clear: what is circular is stating that all "terms" are "abstract
> resources" _hence_ it is an error to dereference.

Sorry, I just don't see that as circular.

Precisely how do you consider a term *not* to be abstract?

And exactly how to you suggest we dereference a URI that denotes
an abstract resource and obtain a representation of that resource
(not a description of or other proxy for that resource)?

You seem to be arguing that the ability to dereference some URI
and get *something* back means that whatever is supposedly
denoted by that URI is not abstract.

Rather, I would say that if you dereference a URI and get something
back, that something is a representation of what that URI *trully*
denotes, and if you thought it denoted an abstract resource, for
which no representation can be obtained, then you were simply
mistaken.

> I've seen nothing to suggest that there is any "overloading" going on.

If a single URI denotes more than one thing, then it is overloaded.

See the attachment.

> I consider it a _fundamental principle_ of the Web architecture that names
> may be dereferenced.

And as I mentioned above, the conflict clearly is between the
Web and the Semantic Web. On the Web, one expects URIs to dereference
to something. On the Semantic Web, one simply expects URIs to name
one and only one thing, whether it is web-accessible or not.

Perhaps the TAG is not actually concerned with the Semantic Web. Perhaps
it is acceptable for them to suggest/encourage namespace names to
dereference to namespace documents. Perhaps we should just abandon
the Semantic Web and just stick with HTTP.

I'd just as soon not.

Patrick
 

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com