Re: XML Schema draft populates the intersection of Language and InformationResource [ISSUE-14 httpRange-14]

On 9/29/07, Richard Cyganiak <richard@cyganiak.de> wrote:
> > You're right, that would be very unfair :).  I'm not blaming you in
> > particular, but the *general* way in which (semiotic) denotation and
> > web architecture's notion of identification are used interchangeably
> > (in more places than just in the RDF specifications).
> > In the long run, the ramifications for this abuse of notation are
> > significant in my opinion but hard to articulate in the short term.
> > They are the result of an impedance between web architecture and KR
> > that is being systematically ignored.
> The distinction you make seems to be too fine for many people
> (including me, unfortunately) to grasp. Care to point me to someplace
> where the mismatch is articulated?

Unfortunately, I don't know of any literature that articulates the
mismatch as a direct comparison of KR versus web architecture
"semantics".  However, if you read the paper "What is Knowledge
Representation" [1] against the backdrop of some of the consensus
points (for which httpRange-14 is one) around conclusions intelligent
agents *should* make by virtue of activity in the transport layer
alone you will find the following points of contention:

"KR is a set of ontological commitments. It is unavoidably so because
of the inevitable imperfections of representations. It is usefully so
because judicious selection of commitments provides the opportunity to
focus attention on aspects of the world we believe to be relevant."
-- Role II: A KR is a Set of Ontological Commitments

If you consider URI identification as equivalent to an 'inert' form
[1] of denotation then that would suggest that webarch (with respect
to 'meaning') considers assertions made about (and within)
webarch:def-information-resource instances as the *only* commitments
relevant to an intelligent agent.  By "within" - I mean the process of
dereferencing a racine and the follow-your-nose-to-an-appropriate-RFC
mechanism you outlined before.

"The third role for a representation is as a fragmentary theory of
intelligent reasoning. This role comes about because the initial
conception of a representation is typically motivated by some insight
indicating how people reason intelligently, or by some belief about
what it means to reason intelligently at all." -- Role III: A KR is a
Fragmentary Theory Of Intelligent Reasoning

Given the inert nature of URI 'denotation', it is clear that web
architecture (even when equipped with the very binary inference
mechanism afforded by httpRange-14) does not satisfy this role by
itself.

"Finally, knowledge representations are also the means by which we
express things about the world, the medium of expression and
communication in which we tell the machine (and perhaps one another)
about the world. This role for representations is inevitable so long
as we need to tell the machine (or other people) about the world, and
so long as we do so by creating and communicating representations." --
Role V: A KR is a Medium of Human Expression

NOTE: I believe the use of the word representation here is completely
orthogonal to web-arch's use of the same word.

If indeed a KR is meant to work as a medium for expression on our
behalf it would have failed with respect to those expressions that are
about things that do not emit codes over a transport protocol.  Before
there were transport protocols, there were expressive languages
(including the one I'm currently communicating to you with), so it
follows that the 'universe of discourse' of web-architecture is a
(very small) subset of those 'things' for which it is useful to
formulate expressions and thus it is not a very useful medium of human
(or machine - for that matter) expression.

This paper is easily my favorite reference on KR as it describes it as
something of an architectural style rather than in terms of a specific
KR (which often loses something in the translation).  In addition, my
running definition of "intelligent agent" is any bit of software that
seeks to take advantage of the roles described in that paper on behalf
of a human.

> Hm. I find it curious that you choose to ignore those parts of the
> relevant specifications that actually address some of the mismatch
> you posit.

Fair enough.  Richard (or any one else for that matter),  Could you
point me to those parts I ignored that address some of the disconnects
I've mentioned above (explcitely)?

> Then Pat is simply asserting that <PatHayes.html> is a member of two
> disjoint classes. OWL and RDF allow us to do such things.

Syntactically, yes we can do such things.  But I was more concerned
about consistency (which is critical for intelligent agents).  In this
case we would have inconsistency between what a (first-class) KR
asserted and the (very weak) assertions about the nature of a referent
which follows from httpRange-14 (if you consider it an ontological
commitment - I certainly don't).

> To spell it out: httpRange-14 says, if it emits representations, then
> its a webarch:informationResource. From AWWW we can conclude that
> webarch:informationResource owl:disjointWith foaf:Person. Thus, if it
> emits representations and claims to be a person, there's a
> contradiction.
> What is your claim? That this is incompatible with rdf-mt?

My claim is that this would be (blatantly) incompatible with the
rdf-mt/OWL-guided interpretation that would follow from the assertions
in the document.  My secondary claim is that this incompatibility is a
strong indicator that web-arch is impersonating a KR at best and the
consequences are dire for future agents.

> As I said before, I *do* believe that RDF's use of URIs should be
> subject to the RFCs governing URI-space. But note that the RFCs place
> no restriction whatsoever on the denotation of large chunks of URI
> space.

Yes, but then we have the following AWWW assertion:

A URI owner SHOULD provide representations of the resource it identifies

Sounds like a restriction on URI denotation to me (at least with
respect to the use of URIs on the hypertext web)

> The most interesting edge cases are http:// URIs that answer 303, and
> URIs with fragment identifier where the racine answers an application/
> rdf+xml representation. They can denote whatever you want, but you
> should set up an associated description that tells the rest of the
> world what you intend to denote. As you know, this has certain
> benefits and certain costs, but those are a separate topic.

What makes these any more 'interesting' than URIs which do not have
representations but which denote referents (in the real world) which
are crucially important to certain domains?  The only 'interesting'
characteristic is that they 'fit' into the web-arch style, but this
should be icing on the cake not a necessary component.

-- Chimezie

Received on Sunday, 30 September 2007 13:02:20 UTC