Re: What the meanging of "is" is

On Wed, Mar 11, 2009 at 7:17 PM, Larry Masinter <masinter@adobe.com> wrote:
> I think the right venue for resolving some of the philosophical questions of
> identity and URIs belongs in the IJCAI conference workshop and not on the
> TAG agenda.

No question. (I don't know whether I can attend, but will look into it.)

> Many TAG and other semantic web discussions about identity and reference ask
> questions about equality, identity, counting, and distinguishing
> “resources”, for example, “ Is the resource identified by URI X the same as
> the resource identified by URI Y?”
>
> “How many resources does a URI identify?”

The question is important in any context where one might consider the
correctness of substituting one URI for another. Such situations do not
necessarily fall under the TAG agenda.

Actually I don't recall these issues ever having come up on TAG time. The TAG
is not really very interested in this level of analysis.

> My perspective is that these questions are not well formed. Equality and
> countability are mathematical operations. The space of “resources” is not a
> set, has no well-defined notion of “identity”, and there is no way, in
> general, of determining “equality”.

I would not say "mathematical" unless you consider any consideration of
equivalence of syntactic things to be mathematical.

But clearly substitutability is a function of interpretation
(semantics), and that is
probably what you mean by both "mathematical" and "context".

> Since you cannot, in general, determine equality of two resources identified
> by two different URIs, counting them or arguing “sameness” doesn’t work.

Agreed that without agreeing first on semantics, the question is meaningless.
But that is true of all questions.

Just because you don't know enough about some things to distinguish them
doesn't mean you shouldn't try to hone your theory so as to resolve the
question - either by fiat, by desired ensuing behavior, or by recognizing
that the answer is inherently unknowable (your theory, or state of knowledge,
is incomplete).

I don't see these questions as being particularly philosophical. With RDF
and its derivatives you are interested in the very practical problem of
automated inference (broadly construed to include search). In order for
inference to be meaningful, you have to choose an inference system and
suitable axioms, sometimes called a logical model. When you interpret the
URIs in some particular way you get a "model" in the opposite sense (that
of model theory, where the problem-domain interpretation is a model of
the theory, not
vice versa). When inference gives you correct results under an interpretation
that is helpful to you, you can declare the whole programme a success
for you.

I think this is all pretty clear, and spelled out in Pat's RDF semantics
document, although perhaps not widely understood.
So I consider these to be much more engineering questions than
philosophical ones.

The thing I would like to contribute is a document (finding?) that
relates HTTP to RDF and puts the httpRange-14 issue to rest so we
can stop bickering about it. I would like to say that there are many
possible theories that might relate the two protocols (stretching a bit
here by calling RDF a "protocol" but I do so to make a point). People who
need such theories (e.g. for the purpose of expressing metadata)
might choose to agree on one or two of them
among themselves so that they can integrate data. Unfortunately
we have at least three such logical models right now that are
mutually inconsistent, and perhaps all are needed. I say: So it goes,
that's the nature of the beast. Different models for different purposes.
As long as you don't simultaneously apply inconsistent models
to the same URI, everything should work out OK.

The httpRange-14 rule can be reinterpreted: If you make an RDF theory
of HTTP or of  some 200-yielding HTTP resource, please try to put
the RDF referent of the URI in classes that are fairly closely tied to the way
HTTP is used - e.g. documents, web pages, REST resources, web services,
and so on. Make "identification" under HTTP as close as you can to
"identification" in a model of your RDF theory.  This is just what some
set of people consider good RDF practice
(for reasons I needn't go into again). "Information resource" does not
have ontological status as a "class", nor does it have any kind of
normative status as establishing a normative HTTP/RDF translation
is not desirable. IR is just a shorthand for certain considerations around
relating the two protocols, improving the chances of consistency
when independent logical models using the same URI collide.

If we could relate HTTP to RDF, this could set an example for relating
other protocol pairs, and if the approach became methodical, we might
be able to make suggestions about some general theory of the
(recommended) meanings of URIs - identifying equivalences, types, and
relations in one domain and showing how they correspond to
equivalences, types, and relations in another.

Personally I think HTTP/RDF is the only case worth pursuing, and
it should be useful to limit the hunt to this one quarry.

Whether anyone else will like this, I'm not sure - certainly the semantic web
view is different from the above, since it thinks the world is just
one big happy
ontology. I don't buy that. (RDF != semantic web)

Feel free to join AWWSW, which is where we talk about this kind of
thing (although
I've put a temporary gag order on discussion of what is an IR).

Received on Thursday, 12 March 2009 13:03:08 UTC