Re: "On the Web" vs "On the Semantic Web"

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Some (all?) of the semantic problems in this thread seem to flutter
around the definitions of terms. I can't always tell, however, if the
problem is that the definitions are contradictory or if some
participants in the discussion simply wish that they were defined
differently.

Consider the following URI:

  http://norman.walsh.name/2003/07/25/attention

In the web architecture, that URI identifies a resource.

Let's leave aside for just a moment the question of what a resource is
and observe that I have a nice label for this one (whatever it is).
Someone, Pat I think, observed recently that the way things are setup
now, if we both use the same identifier (URI), we are trivially
talking about the same thing. That is, there's no mechanism provided
for saying that these two labels that are spelled the same mean two
different things.

So, whatever it is, we can talk about it. We can also hand this label
to software systems and get back representations. For this particular
URI, you can get back HTML, XML, or RDF representations. (I could
provide PDF, PNG, Word Document, and audio representations if I
wished, along with many others.)

The only thing these representations really have in common is that they
are all representations of http://norman.walsh.name/2003/07/25/attention.
That they are representations of the URI I provided is true by definition.

Given that these representations are all peers (that is, each one is as much
a representation of http://norman.walsh.name/2003/07/25/attention as all the
others), it seems to follow that no single one of them can *be* the resource.

So we might conclude that the actual resource is "the abstraction of
the essay I wrote about a hike on Whistler mountain". I'm not sure
there's a better answer. But from there, it seems a few short steps to
use URIs to identify galaxies and trees. All you can ever touch are
the representations that you get back, so what difference does it make
if the underlying thing is abstractly an essay or abstractly a galaxy?

/ pat hayes <phayes@ihmc.us> was heard to say:
[...]
| But do they, in fact? Consider just HTML and http:  URIs for the
| moment, and pretend that the SW was still just a gleam in TimBL's eye.
| It seems to me that RFC2396 *still* doesn't make sense.  It says for
| example that a resource is 'anything with an identity', that each URI
| must identify a single resource, and that (barring network problems,
| etc) the URI enables one to perform operations on the resource. OK,
| take that at face value, and check out the URI
| http://www.ihmc.us/users/phayes/mb#blumlum
| What is the resource identified by that URI?

I don't know. But I can retrieve a representation of it.

| In what possible sense
| could this URI enable me to perform an operation on a resource?

I'm not sure what sort of description you want. That URI does enable
me to perform an operation and I do back a representation. Proof by
example.

| Is
| what I just did impossible, or illegal, or just plain naughty? Or is
| RFC 2396 calling me a liar?

Now I'm really confused. On the one hand I have an identifier with
which I can perform an operation. And when I perform that operation, I
do get back a representation. On the other hand, you seem to be
talking about some assertions made inside the body of the
representation I get back. The fact that the representation contains
some English language text that contains false assertions is
irrelevant.

In short: the assertion that 'There is nothing on or off the World
Wide Web that is denoted by "http://www.ihmc.us/users/phayes/mb#blumlum"'
is demonstrably false. But so what?

I can assert that I am extraterrestrial with three arms and two heads
if I want, that doesn't make it true. The system has to deal with
false assertions.

|>I'm comfortable with electronic proxies for physical things.
|
| So am I, but I would like to be clear when Im talking about the thing
| and when about the proxy; particularly if the proxy in fact is a
| symbol referring to the thing its the proxy of.

Ah, then give the two things distinct URIs and distinguish between
them to your hearts content.

  I assert that http://norman.walsh.name/2003/07/25/attention is the
  the essay that I wrote (abstractly, not one of its representations).
  I assert that uuid:83bde639-7a75-42ac-897c-d34d55fb2b5f is the proxy
  that returns representations of my conceptual artifact.

Now I can distinguish between the two, but I don't often find it
necessary or interesting to do so. I assert that
http://norman.walsh.name/knows/who#norman-walsh identifies my physical
person. As it turns out, you can get either HTML or RDF
representations of me directly over the web. I don't see why it's
important to distinguish between me and the apache server process on
the machine in my closet that's acting as my proxy and returning
representations of me. But if I did, I could give it a distinct name
too.

| I'm only asking y'all to not make
| assertions *about semantics* that don't make semantic sense.

Is there precisely one universal set of semantics? Is it necessary
that if two models describe the world in different ways that (at
least) one of the models must be wrong?

And if it is, does it matter?

On the former question, I'll defer to the philosophers and
semanticists if they tell me they've thought about the problem a lot
and they know the answer.

But I think I can demonstrate that it doesn't always matter. I
routinely rely on devices that operate on the principles of classical,
Newtonian physics. That model is wrong, but the fact that my
speedometer isn't calibrated for time and space distortion effects
caused by my relative motion never matters.

I don't know if that's a useful point to raise or not, but I am very
deeply concerned that we appear to have at hand an issue that is
intractable.

Some folks hold one world view in their heads and argue that the web
works and nothing about their model interferes with their ability to
write software that gets useful work done.

Other folks hold a different world view in their heads and argue that
things are fundamentally ambiguous in ways that are disastrous and
seriously hampers their ability to write software that gets useful
work done.

For the record, I find myself most often in sympathy with the former
group. But I'm using semantic web technologies to get useful work
done, so I must be confused about something.

                                        Be seeing you,
                                          norm

- -- 
Norman.Walsh@Sun.COM    | Time wounds all heels.
XML Standards Architect |
Web Tech. and Standards |
Sun Microsystems, Inc.  | 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.7 <http://mailcrypt.sourceforge.net/>

iD8DBQE/IXZgOyltUcwYWjsRAklWAJ0Z8h+jLaiLHt4YiM/m1X95D6/aNACgrYio
BS4xpdruMwXsnEfKGdH/fjo=
=nvjr
-----END PGP SIGNATURE-----

Received on Friday, 25 July 2003 14:27:00 UTC