Re: Some TAG review of "Cool URIs for the Semantic Web"

>On Thu, 2007-09-20 at 15:59 -0500, Pat Hayes wrote:
>[...]
>>  >>Isn't this 'identifies' usage supposed to suggest SOME kind of
>>  >>similarity between the URI-to-information-resource relationship and
>>  >>the URI-to-Dan Connolly relationship?
>>  >
>>  >Yes, of course.
>>
>>  Well, you can say that the first one is a (very) special case of the
>>  second, with the Web added, I guess. But even that isn't obvious: its
>>  only true by stipulation.
>[...]
>
>I can go with that. "true by stipulation". I'd say that's the
>nature of must/most of Web Architecture. It's true by stipulation...
>or perhaps: it's a way of looking at things.
>
>It seems to me that these 3 situations are completely analogous:
>
>A: 1, 2, and 3 identify numbers
>B: http://xmlns.com/foaf/0.1/Person identifies the class of persons
>C: http://www.w3.org/2002/07/owl#Class identifies the class of classes
>
>In the case of numerals, you're happy to stipulate that they
>identify numbers even though we both know that there are
>multiple theories about how to construct numbers out of sets,
>and they're mutually inconsistent.
>( http://en.wikipedia.org/wiki/Natural_number gives details.)
>And we know that this is not a bug to be fixed, but a fact
>about sufficiently rich formal systems.

Yes, though that point you raise about various 
kinds of set-theoretic construction isn't really 
the issue. The point is that a formal axiomatic 
theory of arithmetic will always have 
non-standard models, ie models which satisfy it 
but which aren't isomorphic to the actual natural 
numbers (which Im happy, like most everyone else, 
to believe that I have a direct intuitive grasp 
of, and which are unique.)

>
>Likewise, for social networking purposes, the FOAF community
>stipulates that foaf:Person is that class of things with
>roughly 6 billion bipeds in it, largely oblivious
>to the DOLCE and EPISTLE refinements of this way of thinking.

Sure. And if FOAF is used ONLY for social 
networking and ONLY isolated from other 
ontologies which talk about people, then fine. 
But is that helpful isolation from ontological 
subtleties in fact going to be maintained? I see 
no way to guarantee it.

>
>And for lots of purposes, http://www.w3.org/2002/07/owl#Class
>is the class of things that stand in the rdf:type relation
>to something (plus the owl:Nothing class), and it doesn't
>matter whether you look at it from the OWL DL or the OWL Full
>formal definition.

For lots, perhaps most, it does not matter. But for some it does.

But my real point is that to be told to think 
about things like the universal set or a unicorn, 
and ask oneself ontological/metaphysical 
questions about those (is this an information 
resource?) in order to decide how to handle a URI 
seems wrong-headed.

>Yes, the WebOnt WG gave two definitions
>for this resource, and the rest of W3C went along with it.
>I think this is somewhat counter to "URIs Identify a Single Resource"
>http://www.w3.org/TR/webarch/#pr-uri-collision but the
>cost of being architecturally pure might have been even higher,
>and we all seem to get along.

We do because everyone in fact uses the DL 
version. Just like everyone with a PC gets along 
with Microsoft :-)

>At least the two definitions
>don't conflict in any way that's observable within the formalisms.

It is observable, but one needs a pretty good 
lab-grade ontological microscope to do the 
observation, I'll agree.

>It's in that sense that I think the TAG is reasonable in
>suggesting this text for the "Cool URIs..." document:
>
>"On the Semantic Web, http: URIs identify not just Web documents, but 
>also real-world objects like people and cars, and even abstract ideas 
>and non-existing things like a mythical unicorn. We call all these 
>things resources."

Sigh. If you can't see how unbelievably SILLY 
this sounds to anyone outside our geeky W3C 
world, I give up. "Resource" is an English word 
with a definite meaning. Not everything is, in 
fact, a resource. If someone were to come up to 
me and say that he had decided to call everything 
a 'foodle', I would think him eccentric. If he 
told me he decided to call everything a 
'pitchfork', I would say he was mad, because many 
things aren't pitchforks. Guess which category 
the above falls into.

>
>[... though I'm somewhat persuaded to take unicorns out of there.]
>
>Now you could continue this argument and say
>that there are similarly two conflicting
>theories about http://www.ihmc.us/users/phayes/PatHayes , one
>that says it's a web page and one that says it's a person, but
>for many purposes, there's no need to look at things that closely.
>The TAG decision on httpRange-14 is that if you look at
>things closely enough to get a 200 response, then it's an
>information resource, and hence not a person. I continue to
>think that's good advice, though I'm fully aware that it's
>not the only coherent rational position to take.

OK, fair enough. But now, lets cut out all this 
gabble about 'information resource'. As I have 
always suspected, this is ALL about HTTP codes. 
If you do a GET with a URI and you get a 200 
response, then the URI is understood to denote 
the resource that sent you that response, a 
(REST-)representation of which you now have. 
Otherwise, you know nothing at all about that the 
URI denotes: it might denote anything.

That is the entire content of the http-range-14 
decision. It's not about "kinds of resource" 
(whatever the hell that can possibly mean) at 
all: it doesn't need the concept of an 
information- or non-information resource to be 
explained, or even for any kind of resource to be 
described, other than the kind that can emit http 
codes and the kind that can't. So why not just 
say this? Its clear, simple, accurate (AFAIKS), 
free of jargon, and avoids all this interminable 
discussion about how to recognize a 
non-information resource when you meet it in the 
street, and damn silly decisions to give a name 
to a dustbin category. We already have a name for 
the only category we need: they are HTTP 
endpoints.

OK, I won't push the ambiguity point for the 
people/web-pages case any more. So as for people, 
we can reasonably assert that people aren't this 
kind of thing: a person can't be an http 
endpoint, so if you get a 200 code back then the 
URI doesn't denote a person. (Though I wonder... 
could we set up a Turing-test/Chinese-Room type 
thought experiment for http? Is there any way one 
could distinguish a human from a machine by 
sending them http?  But lets not go there.)

Pat


>
>
>
>
>[...]
>>  No, that "identify" is one case of "name". Which is true enough for
>>  government work, I will agree. But not all names are identifiers, is
>>  my point.
>
>On that and other matters, I'll perhaps reply separately.
>I'm still thinking it over.
>
>
>--
>Dan Connolly, W3C http://www.w3.org/People/Connolly/


-- 
---------------------------------------------------------------------
IHMC		(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32502			(850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Monday, 24 September 2007 23:06:00 UTC