RE: "tdb" and "duri" URI schemes...

Halloween is over, but I can still raise a specter of an 
indescribable thing -- I can't describe the costume, 
though.

Some things have unique, short descriptions:
short enough to be practical to write within the payload
of a media resource, comprehensible enough for a person
to read, view listen to, make sense of, and interpret,
and unique, in that the description distinguishes the 
thing from every other thing that also has such a 
description.

Not "every" thing can be practically described in such 
a way.

Of course, I can't really tell you which things those are.
But I know that there are infinitely many things which
cannot be described in such a way. 

I can describe "the set of real numbers between 3 and 4, 
inclusive"  and I can describe "the real number pi", but
I know there must be real numbers between 3 and 4 which
do not have short, unique descriptions, since there are
more of the former than there are of the latter.

Perhaps this isn't a big or important point in the definition
of "tdb", but I think it's an interesting one, don't you?

I'll add something about this to the document; I can see
that I need to cover the practical aspects of description
as a means of identification.

Larry
--
http://larry.masinter.net


-----Original Message-----
From: Jonathan Rees [mailto:jar@creativecommons.org] 
Sent: Friday, November 05, 2010 6:50 AM
To: Larry Masinter; Alan Ruttenberg
Cc: W3C TAG
Subject: Re: "tdb" and "duri" URI schemes...

Larry, I think you are a pragmatist. Tell me, what practical ill
consequences are you hoping to stave off by raising the specter of
undescribable things - what bad thing would happen (that does not
happen already) if a URI were used to name or "identify" something
that can't be described?

Why would using a URI to name something that can't be "identified", or
to "identify" something that can't be named or described, be any worse
than anything one might currently (within spec) do with a URI?

Theorem: Everything can be named.  Proof: Suppose that not everything
can be named, i.e. there is something that can't be named. Call is x.
Now x is a name for that thing. Therefore it can be named.
Contradiction, QED.

Theorem: Everything can be described.  Proof: Suppose that not
everything can be described, i.e. there is something that can't be
described. The phrase "can't be described" describes that thing.
Therefore it can be described.  Contradiction, QED.

I don't see the point of even hinting at this sort of gobbledygook, as
"can be described" certainly would.

I think the question is whether to view URIs generally as untyped
names or variables or tokens, in the logical and mathematical (not
set-theoretic, not higher-order) tradition, or whether to see them as
*typed* variables ranging over (as you say) some designated domain -
or perhaps as not involved in naming at all, but only in some special
"identifying" relationship. I have yet to hear a coherent account of
what sorts of things are *not* supposed to be named by URIs, or why,
or even a single example; or how you tell whether a particular thing
has been "identified" by a particular URI.

The untyped approach (logic, RDF) is more parsimonious as it has less
to justify. I know that theories of identification exist that sound
similar to yours (I'm just looking at a book _Pragmatism and
Reference_ by Boersma and it's summarizing Searle's seven rules of
reference, Kripke's cluster theory, etc.) but I don't see how they
apply to the RFCs and the Web - for example, how would either of us
come to know what http://news.google.com/ "identifies", in a way that
would be likely to lead us to agree? How would I falsify
"identification" - prove that http://news.google.com/ and
http://mumble.net/ are different? How do I even know that
http://news.google.com/ "identifies" a "resource" at all?

(If you answer these questions, I will trot out more examples -
random.org, bank session, http://en.wikipedia.org/wiki/Special:Random

- actually not a bad exercise - getting at the other subject I wanted
to write to you about, attribution of time variability.)

Maybe you are saying not that a URI must name something that *can* be
described, but that it must name something that *is* described, if the
URI is to be used?  Maybe you're saying that if you use a URI, you
take on an obligation to be able to come up with a description of the
thing named, on demand, and not just use it while refusing to explain
it? Or maybe you have to be able to "identify" the thing given the
URI, when presented with a police lineup? These rules would relate to
circumstances around the use of a URI, not range constraints on the
thing that the URI names, so at the very least should be phrased
differently. In any case it's not clear that such a rule should be
imposed, or that it would be operational in the sense that we would
all know how to follow it.

Even if we were king and could impose a coherent theory of identity
and identification to all the situations where URIs are already used
(other than RDF, which specifies using them the way first-order logic
uses variable and relation symbols - bypassing any demand for
"identification"), I can't see what purpose it would serve, and it
would probably end up being so complicated that it would do more harm
than good.

As usual this may sound rhetorical but I'm actually interested in your answers.

Jonathan

(once again conflicted over whether to acknowledge Pat Hayes, since I
do not want him to be blamed for all I got wrong!)

On Fri, Nov 5, 2010 at 4:38 AM, Alan Ruttenberg
<alanruttenberg@gmail.com> wrote:
> On Fri, Nov 5, 2010 at 4:13 AM, Larry Masinter <masinter@adobe.com> wrote:
>>   FWIW, I can't understand what you are talking about here.
>>
>> you seem to, or at least your argument seems to accept the terminology
>> anyway.
>
> I'm doing my best to communicate. That I accept terms you use as
> meaningful doesn't mean I accept the sentences you make from them as
> such.
>
>>   Even if there was a counting argument to be made here, I can't see how
>> you would arrive at "things that *can* be described". You might perhaps land
>> up with "things that *have* been described". But there isn't a unique
>> mapping of countable on to uncountable sets.
>>
>> I don’t really think it’s necessary to get into a debate about determinism,
>> I’m just trying to justify my choice of wording here. I say “things that can
>> be described”, and Jonathan said it is “silly and meaningless”. I disagree
>> that the distinction is either “silly” or “meaningless”, even if it doesn’t
>> match your (or his) world model.
>
> It's "silly" because it doesn't have a constructive impact and because
> it suggests a bad thinking habit - namely that there are two
> categories of things, somehow intrinsically different - those that can
> and those that can't be described. Whether you intend this
> interpretation or not, this is what the language you use suggests.
>
> Whether it is meaningful is still unclear to me. We don't have to
> debate determinism, but you should be consistent in your language. If
> your argument depends on determinism then call the set "things that
> will be described". If not, explain what you mean by "can".
>
>>  At least, for me, the distinction isn't silly or meaningless. In addition,
>> the notion of "identity" is
>> associated with the description rather than the thing-described
>>
>>  Which notion of identity? There are a
>> number. http://plato.stanford.edu/entries/identity/

>>
>>   There are certainly notions of identity associated with (any)things.
>>
>> Perhaps you can explain how I can talk about identity of things without
>> having a way of talking about the things whose identity is in question.
>
> Are we not able to talk about the identity of real numbers, despite
> the fact that there are uncountably many, and therefore by your
> account we can't possibly describe all of them?
>
> (answer: yes)
>
>>   That is, "things" don't really form a set, in the sense of having a clear
>> equality relationship.
>>
>> Set theory requires you to know that the elements of a set are distinct.
>> Either a = b or a != b.  Unless you can identify a or b, you can’t even ask
>> whether they are the same.
>
> Set theory doesn't ask whether elements are the same. It doesn't ask
> you to identify anything. There are things. There can be sets of them.
> The structure of the set is such that no thing appears more than once.
>
> We are perfectly able to create the set of real numbers in the
> interval [0,1], despite that in the framework you present we don't
> have enough descriptions (which are countable by your assertion) to
> "identify" all the a's and b's.
>
>> I'd be interested in an attempt to be convinced. But as another heads up,
>> what you are saying here seems in contradiction to the basis of all the
>> SemWeb languages, which I think would be setting precedent.
>>
>> I can’t imagine what this means. If you have something in specific you think
>> I’m contradicting, please point it out.
>
> http://www.w3.org/TR/rdf-mt/#urisandlit

>
> "The semantics treats all RDF names as expressions which denote. The
> things denoted are called 'resources', following [RFC 2396], but no
> assumptions are made here about the nature of resources; 'resource' is
> treated here as synonymous with 'entity', i.e. as a generic term for
> anything in the universe of discourse."
>
> http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/#Real_Numbers.2C_Decimal_Numbers.2C_and_Integers

>
> "The datatypes owl:real and owl:rational are defined as follows.
>
> Value Spaces.
>
> The value space of owl:real is the set of all real numbers.
> The value space of owl:rational is the set of all rational numbers. It
> is a subset of the value space of owl:real, and it contains the value
> space of xsd:decimal (and thus of all xsd: numeric datatypes listed
> above as well)."
>
> So the universe of discourse includes real numbers, which are
> uncountable. Your restriction of the domain of discourse to any
> countable set ("things that can be described" distinguished by being
> countable) contradicts this precedent.
>
> -Alan
>

Received on Saturday, 6 November 2010 06:23:47 UTC