W3C home > Mailing lists > Public > www-tag@w3.org > October 2004

Re: [Fwd: RE: "information resource"]

From: Harry Halpin <hhalpin@ibiblio.org>
Date: Fri, 15 Oct 2004 02:27:35 -0400 (EDT)
To: www-tag@w3.org
Message-ID: <Pine.LNX.4.44.0410150155130.6819-100000@tribal.metalab.unc.edu>

This notion of "information resource" is definitely tricky, but I concur
as an outside observer that TAG is on the right track. 

I'm still not sure if this is really testable in all cases, but it
seems to lie on some pretty good common-sense intuitions. I guess
the test would be "If this *thing* is represented, is the representation
still *this thing*, and not a *representation of this this thing*"? If you 
take a picture of a dog, you clearly a "representation of a dog", not
*an actual dog*. If you have my medical record and take it from a piece of
paper and put in into some database, then general an XML file from that 
database that flies across the Web, that XML file is *my medical record*
in the same sense that the original paper version is and not *a 
representation of my medical record*. In fact, the word *representation
of my medical record* doesn't even make much sense, does it?  

By hook or by crook matters of reference are being dragged into the TAG. 
>From the current definition of "information resource", all  "essential 
properties" are conveyed in the digital message.  The question
is then "what is essential?" A real life shaggy dog has many essential 
properties that are resistant to being made part of a digital 
representation, such as a smell. However, is the smell "essential", and
is that adjective by nature not subjective? Who dictates what is 
essential?

Any abstraction of my real life shaggy dog would doubtless lead many 
properties behind that *someone* could judge to be essential. But in 
domain of medical records, there are standard things that are measured 
(like heartbeat) and written down, and *all of this information* can be 
conveyed in message (in XML, RDF, binary). Things like the type of paper 
the record is on or the expression on the doctors face as they wrote 
down the heartbeat are inessential, because they are judged not to be 
part of the medical record. The reason all a medical record can be conveyed as 
an information resource while a dog cannot is because a medical record
is *already abstract information* while a real-life shaggy dog is *not
abstract* at all. The key notion seems to be a *level of abstraction*,
and it certainly seems somethings that are abstract (like numbers) are
much more amendable to being digital (and thus being conveyed on the Web)
than things like smell or the gleam in someone's eye.

Now, there are border-line cases - what about  "the class of all tigers?" 
It's abstract - but what are it's essential properties? Being a feline and 
a quadriped? But as Searle notes, a tiger with a leg missing is after all, still a 
tiger. It appears the idea of the "class of all tigers" does not exist
at suitable level of abstraction for description of its essential 
properites, much less digitization of its properties, - and so the
"class of all tigers" is not an informaton resource.  Correct? I mean,
where's the test? 

So, it appears an information resource is something whose essential 
properties can be conveyed, as a message, because the thing being conveyed
exists at a level of abstraction where it can be digitized in its 
entirety. This has ramifications since such a thing can be realized in 
multiple formats. Your real-life dog can only be represented by that thing 
that *actually* smells, the class of all dogs can only be represented by
the actual class of all dogs (existing, existed, and perhaps to exist or
imagined) and also can't be a resource. But an instance of a medical 
record, being basically numbers and strings, is just the right level of 
abstraction and digitization to be an information resource. Am I correct? 

I might add it appears we're all using "information" in a Dretskian sense, 
not a Shannon sense. If I get a spare day or two in the next week I'll
write a quick squib on various stabs people have had at defining
information, and post it to both www-tag and xml-dev since this
notion keeps returning. 

I would recommend reading D. Israel's "What is Information" and
F. Dretske's "Precis to Knowledge and the Flow of Information" for
further, and perhaps useful, background information. 

			My two cents,
				Harry 
 

On Fri, 15 Oct 2004, Chris Lilley wrote:

> 
> On Thursday, October 14, 2004, 9:51:39 PM, noah wrote:
> 
> nuic> Perhaps it's worth noting that our current editors draft says [1]
> 
> nuic> TRUE per Basel Definition:  "The distinguishing
> nuic> characteristic of these [information] resources is that 
> nuic> all of their essential characteristics can be conveyed in a message."
> 
> nuic> but it does NOT say the converse:
> 
> nuic> FALSE per Basel Definition: "A non-information resource is distinguished
> nuic> by the fact that none of its state can be conveyed in a message."
> 
> Yes, exactly. And that is good and I am comfortable with it.
> 
> nuic> We shouldn't be surprised that there is some machine-representable state
> nuic> for a real live shaggy dog.
> 
> 
> I'm not. But (see earlier threads on testability) saying that an IR is
> any resource for which there exists any information whatsoever is
> equivalent to saying IR == resource, thus making it a useless and
> meaningless term.
> 
> nuic>   We might choose to expose its temperature or
> nuic> its weight, for example.  The distinction drawn in Basel is that dogs are
> nuic> interestingly different from information resources because there exist
> nuic> essential aspects of the dog that are not conveyable in a machine-readable
> nuic> way.
> 
> Yes and this was a distinction that I was happy to see made, and stopped
> opposing the term IR because it was now both useful and testable.
> 
> nuic> Noah
> 
> nuic> [1] http://www.w3.org/2001/tag/webarch/#id-resources (I don't think I have
> nuic> a stable link in date space for this, unfortunately)
> 
> nuic> --------------------------------------
> nuic> Noah Mendelsohn 
> nuic> IBM Corporation
> nuic> One Rogers Street
> nuic> Cambridge, MA 02142
> nuic> 1-617-693-4036
> nuic> --------------------------------------
> 
> 
> 
> 
> 
> 
> 
> 


	Harry Halpin
	Informatics, University of Edinburgh 	
        http://www.ibiblio.org/hhalpin
Received on Friday, 15 October 2004 06:27:36 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:32:43 UTC