W3C home > Mailing lists > Public > www-rdf-interest@w3.org > April 2002

Re: Documents, Cars, Hills, and Valleys

From: Nick Matsakis <matsakis@mit.edu>
Date: Wed, 10 Apr 2002 23:54:49 -0400 (EDT)
To: Mark Nottingham <mnot@mnot.net>
cc: www-rdf-interest@w3.org
Message-ID: <Pine.OSX.4.44.0204102320550.28520-100000@artoo.ai.mit.edu>

On Wed, 10 Apr 2002, Mark Nottingham wrote:

> Of course. However, the documents/representations/instances are not
> identified by the URI; the resource is.

http GET gives you some bits.  Bits are not printers or people or cars,
but they provide information about such resources. If you use the http URL
to refer to the resource, how do you use to refer to the bits?

The bits have a set of properties of their own that are distinct from the
properties of the resource. How do you propose to make RDF statements
about the bits while preserving the distinction between the representation
and the resource?  Cars have color, while bits have none.  Bits can have a
checksum, while cars do not.

> When you see a car, you're perceiving light bouncing off it; however,
> when you say "that car," you mean the car, not the light.

This leads us into some deep philosophy. while I like the argument I've
outlined above, it has a large hole that troubles me. The hole is the
problem of infinite regression in perception.  Above, I tacitly imply that
the returned by the http GET are somehow primordial.  This clearly isn't
true.

What is to stop me from saying, "Are you mad?  What makes you think you
can get a mathematical object like bits? All you get are java
datatypes/virtual memory addresses/physical memory addresses/voltages in
DRAM/Quantum Wave functions.... you just see the bits *through* this
medium."

In general, there are statements we may wish to make about the physical
memory addresses that don't apply to the bits, just as there are
statements we might want to make about the bits which don't apply to the
document (or car, if you must).

It seems to me that the decisions to stop the regression is up to the
ontologist. Different problem domains require different levels of
granularity. For many computer applications, though, it seems wise to
maintain a firm distinction between resources that can be serialized into
bits and those that can't.  Just think, once we invent transporter
technology, we won't have to worry about this problem any more!

Nick Matsakis
Received on Wednesday, 10 April 2002 23:54:51 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:53 GMT