RE: [httpRange-14] What do HTTP URIs Identify?

> From: Joshua Allen [mailto:joshuaa@microsoft.com] 
>
> Ambiguity is inevitable.  OK?  Is everyone happy now that we 
> agree on this?

I've agreed with this for as long as I've been interested in RDF and the
semantic web. It's been far from obvious to me that you've agreed with
it over the same period, until you started using the word 'confidence'
recently. If I've been asleep at the wheel or I'm misrepresenting your
position, please correct me.

And let's narrow down what is meant by ambiguity. Ambiguity is a one to
many relationship between the URIs and the Resources. That is, there can
be state of affairs where a URI is being used to refer to more than one
Resource. I claim such states of affairs are possible and that the
axioms of the web architecture as outlined by Tim do not acknowledge
this. I further claim the RDF Model Theory was generated precisely to
address such states of affairs. 

Now, why is it we don't have an axiom of ambiguity? It's a fundamental
constraint to the semantic web  in the same way latency is to the
internet. Instead what we have are the axioms you've been throwing about
recently, which do not seem adequate. I'd rather not see the SemWeb
built and them have a Jim Waldo come along telling us we need to throw
out certain (obviously wrong) assumptions.

> But I am puzzled how this has ANY bearing on the following "axiom":
>
> "If two people independently use the same URI as an 
> identifier, they should be able to have a reasonable degree 
> of confidence that they are identifying the same resource.  

Well, finally we cut to the chase; "a reasonable degree of confidence".
Thus we don't need the two URI axioms you've been evangelising recently
as the key axioms; only half jokingly, we need the axioms of probability
or an acknowledgement of them as being truthful.

For the record: I fully agree with Tim's URI axioms, but only as
principles. They are in fact principles I'm evangelising strongly in my
current work. I just don't see them as either obvious truths (which is
what is expected of axioms) or desirable architecture (the coordination
costs are too high, as per backlinking) in a planetary scale information
system along the lines of the semantic web.

 
> People should not be required to parse, dereference, or 
> otherwise acquire any *additional* disambiguating information 
> to provide this basic guarantee.  

First, that's not possible; coordination requires some form of
communication (I'll accept shared indexicals as communication here).
Second, 'should not' is no basis for interop; invert the sentence and
use 'may' instead. Third, it contradicts your first paragraph. 
 

> Resource naming practices should be considered carefully, and 
> people are strongly discouraged from naming resources in a 
> manner that unnecessarily weakens this guarantee."

But there is no guarantee. Degrees of confidence rule that out. Is this
not the case?

regards,
Bill de hÓra

..
Propylon
www.propylon.com

Received on Friday, 2 August 2002 16:01:30 UTC