Re: URIs: resources and contradictions was: Re: httpRange proposed text

On Sun, Aug 04, 2002 at 01:25:24PM +0100, Bill de hÓra wrote:
> > On Behalf Of Jonathan Borden
> > Yes I would say that every time we use a single URI we refer 
> > to the same resource, but perhaps in different contexts. The 
> > meaning of a URI is the same as the meaning of the resource 
> > it refers to.
> 
> We solve the problem of the one to many mapping between URIs and
> Resource by axiomatising that such a relationship is never the case
> (I'll accept that for now as being an internally consistent view in the
> web, but observe it's trivially not true when we relate the world to the
> web). Granted, what something means is often more relevant than what
> something is. 

But the issue is where in the architecture you want to start making
assertions about "meaning" in order to relate the 'real world' 
to the 'web'. Please don't attempt to do it to URIs themselves. Layering
is a _very_ important concept that works. There are systems that
use URIs that intentionally have nothing to do with REST, SOAP, HTTP,
documents or hyperlinks. LDAP, instant messaging protocols, Java RMI, etc
do not have concepts of documents, BASEs, etc and would have trememdous
problems if you started mucking around with their internal assumptions
about uniqueness and equivalence.

> What is left ungrounded then is the meaning of a URI. This can be
> grounded either by a model theoretic interpretation (RDF), the history
> of representations (REST), or some other technique.

The meaning of a URI by itself requires that 'meaning' be left undefined.
That is a concept that is restricted to the systems that _use_ URIs. You
can't impose one on everything else. If you need to talk about the
meaning of URI please do so within the context of some system (RDF, REST,
SOAP, whatever). Please don't change the underlying physics of the entire
URI universe to suit this particular application....

> > A resource may have different meanings in 
> > different contexts (e.g. as asserted by two different 
> > individuals). 
> 
> It would be more consistent with your argument to say that a URI may
> have different meanings in different contexts; this is by your and
> others insistence that a URI point to one and only one thing
> (irregardless of what a group of parties might think that URI actually
> points to).
> 
> I'll also observe that this notion of context dependence disposes of
> axiom 1. But so does the RDF Model Theory, which also disposes of 2a. By
> dispose I mean they are downgraded in RDF from truths to requirements.

Again, what layer are you talking about?

> Under a setup where resource are synonymous with URIs, I see no point in
> having resources, symbols and interpretations alone should be sufficient
> for our computers. As Chris Lilley said about resources, they're tricky
> to get a hold of; while I take issue with one aspect of Chris' analogy,
> physicists do invent particles to explain formulae. In kind, we seem to
> need resources only to explain the existence of URIs. Yet resources seem
> more like phlogiston than neutrinos.  



> I would also expect to alienate many on comers when we tell them, quite
> literally, they don't know what they are talking about.

Given no other knowledge than the URI, yes, they can't really say much.
Its only within a given system/context can they start making 'meaningfull'
assertions. And that's intentional....

> And my main objection is not that it is less right to make URIs and
> resources synonymous, or that it is not aesthetically appealing to me,
> but that it leaves something important unsaid. By not axiomatising
> ambiguity we leave it a ghost in the system; perhaps by claiming that it
> is not there (an untruth), or claiming that it is 'obvious'; as we all
> know it's there it needs no further treatment. The benefits of certainty
> are so many, it's hard to let it go. 

The point is not that its being left unsaid. Its that those things 
exist at another layer/component within the architecture. Its the
same way that IP doesn't say anything about session contexts. TCP is by
and large the most used Layer 4 protocol but just because its popular doesn't
mean we replace IP with it....

The hour-glass model is a useful thing. Don't try and expand the middle bit
to much or you end up breaking the entire thing....

-MM

-- 
--------------------------------------------------------------------------------
Michael Mealling	|      Vote Libertarian!       | urn:pin:1
michael@neonym.net      |                              | http://www.neonym.net

Received on Sunday, 4 August 2002 11:28:50 UTC