Re: CG: XML Conceptual Graphs: an open design?

Hi Murray

Well I should have been clearer.  I meant to say that you cannot ~transmit~
the identity of a thing by transmitting it's URI or any other syntactic
construct whatsoever.  So defining these globally unique strings does not
give us some magical identity matrix in the ether on which we can map our
things.   I think a lot of semantic web builders are (perhaps
subconsciously) thinking, or wishing that it will.

For example:

[1] http://www.ietf.org/rfc/rfc2396.txt  starts with the assertion

     "Uniform Resource Identifiers (URI) provide a simple
      and extensible means for identifying a resource."

It then goes on to say that:

     "A resource can be anything that has identity."

Well I claim that a URI does NOT provide a means for identifying anything
that has identity.  Such a means of identifying things would need to be a
method (a process) acted upon by a community of usage.  Without the method,
without the process, without the community, these globally unique strings
will just hang there against a background of static;  and will not even
constitute information.

Let me continue with your example URI ...

[2]   http://www.altheim.com/xcg/animals.xcg#jackrussell

... well I didn't know what a "Jack Russell terrier" was (thanks for the
great example!); and by transmitting to me your URI, you did not get me one
bit closer.  Rather, since I was curious, I first tried to dereference your
URI and of course, that was useless.  So instead I took your ~description~
and pasted it into google and  after a moment's surfing arrived at a *URL*
...

[3] http://www.akc.org/breeds/recbreeds/jrt.cfm

.. which does in fact, (1) combined with my mental faculties, to effectively
transmit the identity of this thing.   Note: I used your description,  and
(2)  a method which involved a dereference to access the (3) active
community of dog breeders.  Lacking any of those elements (1,2, or 3) your
means of transmitting this unknown identity to me would have failed.

But I agree with you that this "certainly shouldn't stop us from using
existing identification schemes within existing communities" ... nor should
it hold us back from building the processes and communities which actually
will provide us  with "a simple and extensible means for identifying"
anything.

Seth Russell

----- in response to Original Message -----
From: "Murray Altheim" <murray.altheim@sun.com>
To: <cg@cs.uah.edu>


> Seth Russell wrote:
> >
> > > > > [Star: #MorningStar]
> > > > > [Star: #EveningStar]
> >
> > The identity of a thing is NOT it's name.  I doubt there is ever going
to be
> > a way to map the identity of something to a string (regardless of how
many
> > syntactic tricks we invent).  Rather we could grok that ~internal~ to
any
> > system (that cares about identity)  there is a mapping of things that
have
> > identity with the internal nodes in the system.  Arriving at that
mapping is
> > a process.    There is, imho, no syntx only sloution.   We need to work
on
> > the process.
>
> Actually there is a syntax solution, if you work backwards. If you
> create a <concept> element having an ID (identifier) in an XML Conceptual
> Graphs document, post it to a fixed location on the Web, you've created
> a URL that is essentially a binding point for anything that has the same
> identity as whatever that concept is designated to describe. That URL in
> the XTM space is called a Published Subject Indicator (PSI). Any other
> document that wants to associate the same concept with the existing
> one (i.e., "create an identity with") can use the PSI (which could be
> considered a locator but doesn't need to be) as its identifier. So
> for example, if I wanted to talk about Jack Russell terriers, I could
> post something establishing
>
>    http://www.altheim.com/xcg/animals.xcg#jackrussell
>
> and then anyone else who wanted to talk with me about Jack Russell
> terriers could use that identifier. Now, you ask, why would anyone
> want to use Murray's identifier? Well, you might not, but if I
> happened to represent some authority in the dog world (such as the
> AKC) there'd be some weight behind it. Any community can have their
> own de facto or de jure authority, but if it's just me and Peter
> sharing information we can still do it, very egalitarian-like.
>
> So in a nutshell, I agree that absent an authoritarian ruler that
> forces us to all use truly canonical identifiers, there can be no
> universal identifiers. But that certainly shouldn't stop us from
> using existing identification schemes within existing communities,
> such as ISBN numbers, employee identification numbers, or the more
> complex methodology of using names within controlled contexts (such
> as my name within my family, the UN or ISO names of countries in
> English or French, etc.). It's a thorny problem, but not without
> solution.
>
> Murray

Received on Saturday, 25 August 2001 13:11:30 UTC