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

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

From: Dan Connolly <connolly@w3.org>
Date: Thu, 20 Sep 2007 19:23:29 +0100
Message-ID: <46F2BAA1.9030704@w3.org>
To: Pat Hayes <phayes@ihmc.us>
Cc: Tim Berners-Lee <timbl@w3.org>, Technical Architecture Group WG <www-tag@w3.org>, Susie Stephens <susie.stephens@gmail.com>

My main point is about numbers and numerals, near
the end; I hope you'll read that far before you argue any of the
lesser points...

Pat Hayes wrote:
>> Pat Hayes wrote:
>>>> 7. "On the Semantic Web, 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."
>>>> Of course many people would consider the Traditional Web to include 
>>>> mailto: links. Suggest:
>>>> "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."
>>> This seems to me to embody the central error which is causing so much 
>>> trouble. In what sense can a URI (or indeed any name: the fact that 
>>> is a URI is irrelevant in this case) be said to "identify" a 
>>> real-world or nonexistent entity? The direct answer is, it CANT. To 
>>> even use this word "identify" in this sense and in this kind of a 
>>> case, is clearly and provably WRONG.
>> It seems pretty clear that the name "Dan Connolly" identifies
>> me and that I am a real-world entity.
> You are real, I grant you. And "Dan Connolly" is your name. But does 
> that name "identify" you?  Seems to me that in order to justify this
> usage, it would have to be that if someone who didn't know you or 
> anything about you were given just that name, "Dan Connolly", a string 
> of 12 characters, that they could figure out from that string which 
> real-world entity you actually are. Maybe not *locate* you, but at least 
> single you out from all the the other DCs there might be. But in fact 
> they couldn't, because there are others with the same name, who 
> therefore it also 'identifies' if it identifies you. (Eg 
> http://www.hospitalitynet.org/news/4012693.search?query=+%22dan+connolly%22) 
> Is that enough of a proof?

It helps me understand what you mean, yes.

Your point here is about scope. URIs are intended to
have universal/global scope, so it's less clear that this form
of argument applies, though perhaps it still does... in any case...

> Maybe this criterion is set too high, although I kind of transcribed it 
> from the Web requirements for the "identifies" relationship between URIs 
> and 'information resources'. But if this is too high, why do you and the 
> TAG insist on saying "<name> identifies <thing>" instead of the more 
> usual and completely non-puzzling "<name> is a name of <thing>" or even 
> "<name> refers to <thing>" ?

I have a hard time imagining that changing "identifies" to "refers"
or even "is a name of" would make much difference to the audience
of this document.

'name' and 'identify' are synonyms, according to thesaurus.com.
That source doesn't give 'refer' and 'identify' as direct synonyms,
but it says 'indicate' is a synonym of 'refer' and 'identify' is
a synonym of 'indicate'.

Given that it's pretty much a coin flip at this level, I think
it's reasonable to use "identify" in order to help introduce
the reader to the standard "Uniform Resource Identifier"

> 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.

> We usually say 
> "identifies" rather than merely "is a name of" when the identifier can 
> be USED to single out the thing named from the universal multitude,
 > just as an xxtp: URI can be USED to actually GET something.

That's not at all clear to me. What well-known
works make that distinction?

A few seconds of searching yields evidence to the contrary,
i.e. that "name" and "identify" are synonyms:

"Identifiers (IDs) are lexical tokens that name entities."
  -- http://en.wikipedia.org/wiki/Identifier

> But names can't 
> be used like this, in general: they are just names. They aren't anything 
> like addresses or identifiers, because they don't identify. They just name.

That's argument by assertion; I don't find it persuasive at all.

(I could give a counter-argument, but I'm sure you a have
seen it before and it didn't convince you then, so it's
probably not worth going into again now.
In case anybody is new to this conversation,
see http://www.w3.org/DesignIssues/NameMyth for the position
that I hold.)

>> If you have a proof that this is not so, I'm quite curious to see it.
> The proof I had in mind was the fact that just about any large 
> first-order theory has nonstandard models, which follows from Goedel's 
> theorem, for arithmetic,

OK, yes, that's a proof, but it's a pretty obscure point; I wonder
if it is really important for the purpose of this document, i.e. setting
up web servers and such...

> but informally (OK, its not a PROOF any more) 
> is clearly true for almost any reasonably expressive theory with 
> infinite models. Under these circumstances, no amount of ontological 
> content or description is going to be enough to single out one sense of 
> "Dan Connolly" from some others; and there will always be others. So 
> apart from the above objection, and even if one buys into the fantasy of 
> URIs being a globally unique naming system, there will still be SW 
> issues that one ontology thinks of persons like you in one way and 
> another ontology thinks of them in a different way. Im not sure how you 
> like to describe this inevitable situation, but one way that makes 
> semantic sense is to say that there are in fact two (and more) distinct 
> entities, sorry resources, which are both 'you'.  And its no defense to
 > say that you FEEL like one thing. Of course we all feel like that: it
 > only follows that this is what it feels like to be a whole lot of
 > person-things at the same time.

Yes, that point is well made, regarding formal systems.

I suppose web server configuration files are, at some level,
formal systems, but it seems to me that we gloss over the
point about nonstandard models quite cost-effectively
for lots of engineering purposes.

For typical engineering purposes, we agree what 1, 2, and 3
identify, no?

In a conversation with, say, a Java programmer,
are you uncomfortable with saying that numerals identify numbers?
If not, what would you say in that case?

It seems to me that lots of programmers buy into the "fantasy"
that numerals uniquely identify numbers, so the "one buys into
the fantasy of  URIs being a globally unique naming system"
situation isn't that much of a stretch, for the purpose
of this "Cool URIs..." document.

Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Thursday, 20 September 2007 18:23:13 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:17 UTC