W3C home > Mailing lists > Public > www-tag@w3.org > June 2007

Re: Terminology (was Re: article on URIs, is this material that can be used by the)

From: Pat Hayes <phayes@ihmc.us>
Date: Thu, 28 Jun 2007 17:36:00 -0500
Message-Id: <p06230917c2a9d781e632@[10.100.0.24]>
To: Dan Connolly <connolly@w3.org>
Cc: noah_mendelsohn@us.ibm.com, Dan Brickley <danbri@danbri.org>, "Henry S. Thompson" <ht@inf.ed.ac.uk>, Tim Berners-Lee <timbl@w3.org>, www-tag@w3.org

>On Wed, 2007-06-27 at 16:53 -0500, Pat Hayes wrote:
>[...]
>>  >Re "in general," I wonder if you're reading a forall quantifier
>>  >that isn't there.
>>
>>  BUt it is there.
>
>More argument by assertion. This is really no fun.
>
>Thinking this over, I perhaps overstated my case a bit.
>That reference and access are separate has always been
>my position but for much of the development of webarch,
>the TAG wasn't completely convinced of the possibility
>of resources beyond information resources. And certainly
>much ordinary web stuff deals with information resources,
>and most of /TR/webarch is about information resources.

OK, but Im still puzzled by your thinking here. 
What has the existence of any kind of resource 
got to do with the distinction between reference 
and access? Even if all OWL ontologies were 
required to only be about information resources, 
the distinction between reference and access 
would still be germane. And to say that a URI 
'identifies' a resource would still be just as 
ambiguous between these two meanings.

>But please, Pat, if we're to make any progress,
>I'd like you to take care to give supporting argument
>when making a point.

I thought I had cited earlier text in the 
document which (the way I read it) seemed to 
clearly and unequivocally require that the text 
we were arguing about above must be understood as 
universal.

>The text you quoted is:
>
>>>   "URIs are divided into schemes (2.4) that
>>>   define, via their scheme specification, the
>>>   mechanism by which scheme-specific identifiers
>>>   are associated with resources."
>
>Surely it's clear that "identifiers" is existentially
>quantified (for some identifiers...) and not universally
>quantified.

Actually, no, I didn't read it that way.

>The string "qrfl:abc" is a URI but the
>scheme qrfl: isn't registered, so there isn't a
>specification that associates it with a resource.

No, but you know that 'qrfl' has to be a scheme, 
right? Of course some of the machinery might be 
missing, I understand that.

>Likewise, http://example/abc is a URI, but
>if you follow your nose thru the specs, you'll
>see that IANA reserves the "example" domain in such
>a way that doesn't associate a resource with http://example/abc .
>I own http://dm93.org/abc123 and I can say authoritatively
>that I have not associated it with any resource.
>
>
>And the argument by assertion continues...
>
>>  >It doesn't say "the design of the web guarantees
>>  >that a URI identifies one resource"; rather,
>>  >it's saying that the intended design of URIs
>>  >is that each one identifies one resource; if you
>>  >use them some other way, then you're not using
>>  >them as designed.
>>
>>  But you CANNOT POSSIBLY use them that way (if all
>>  you have available is descriptions of
>>  them,anyway, which is the usual case when using
>>  referring names.) So this seems like damn silly
>>  advice/design.
>
>Oh? It seems to me that we do use them that way,
>by the millions, daily. Proper names work similarly;
>by design, Pat Hayes refers to just one thing.

No, it doesn't. Ignoring the fact that there are 
any number of Pat Hayes's, many of them pictured 
on the Web, and say we somehow agree to be 
talking about me; it STILL doesn't, since I can 
be individuated in all kinds of ways. For 
example, some people take "Pat Hayes" to refer to 
me *now*; others take it to refer to me 
throughout my lifetime. Others, such as I suspect 
yourself, don't care about the distinction and 
don't care to make it. Fair enough: but then you 
have to face up to the fact (and it is a fact) 
that there are at least three distinct ways to 
understand what it is that my name refers to, 
even when you successfully use it to refer to the 
person typing this reply, because there are 
several different notions (at least three) of 
what KIND of a thing a person typing a reply IS.

Now, this would just be philosophical 
hair-splitting if it weren't for the presence of 
the semantic web, and the W3C requirement that 
identifiers in Web ontology languages must be 
URIrefs. But those hair-splitting distinctions 
become more like road barriers when you start 
using names in formal ontologies. If you don't 
believe me, just read the recent traffic on 
public-semweb-lifesci@w3.org or 
ontolog-forum@ontolog.cim3.net. At almost exactly 
the same time that Webarch is being comfortably 
sloppy about what names denote while laying down 
cast-in-stone Rules of good Web behavior, the 
most active area in applied real-world Semweb 
ontology development is wrestling with the 
differences between continuants and occurrents.

If you want to say that having a URI refer to me 
is being referentially unambiguous, then give up 
on the idea that what you are saying is going to 
have any application to the real semantic web. On 
the other hand, if you want what you are saying 
to apply to the real semantic web, then PLEASE 
say things that make semantic sense. A person is 
one unique thing only in the context of an agreed 
ontological framework. owl:sameAs written between 
my name in a DOLCE-based ontology and my name in 
a PSL-based ontology is going to rapidly generate 
fatal contradictions, because I'm one kind of 
thing in one of them and a completely different 
kind of thing in the other. And they are BOTH 
RIGHT.

>As you have pointed out, there are lots
>of interpretations that are consistent with
>all the texts I have ever read
>using that name. But by convention, we
>give distinct names to siblings to avoid
>having Pat Hayes refer to two different
>but nearby people

Hah. I have the same name as my grandfather, and 
one of my extended family's homes had three 
'Terry Hayes' in it for years.

>... or we add Jr/Sr suffixes.
>What "By design a URI identifies one resource"
>is saying about URIs is that
>not only should you not give you and your
>sibling the same URI, you shouldn't
>give a city and a person the same URI
>the way we sometimes do with proper
>names such as Lincoln.

But that isn't what the document actually SAYS. 
What it says is that by design, the referent of a 
URI should be *unique*. And that is a silly thing 
to say in a context that includes formal 
ontologies written using URIs as identifiers, 
since if understood to mean what it in fact says, 
it cannot possibly be satisfied. Even if we did 
manage to satisfy it for a while for some URI, 
there is no way to know that some new 
distinction, as yet not thought of, will suddenly 
reveal a previously unsuspected ambiguity (say, 
between two kinds of continuant) that will make 
the URI ambiguous again.

I know that y'all don't want to even think about 
stuff like this, and that (as Tim once said) I'm 
like a quantum theorist who keeps complaining 
about a document written for engineers. But my 
point is that since part of the Web is now 
semantic, and since y'all are using semantic 
language here, that you should at least be aware 
how your words might be misunderstood by the 
quantum theorists who are actually doing some of 
the engineering these days.

Pat

>
>
>
>--
>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 Thursday, 28 June 2007 22:36:14 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:52 UTC