URIs: Identifiers in name, but...

At 10:06 AM 2001-05-01 -0400, Michael Mealling wrote:
>
>A URI identifies one and only one abstract Resource (note the
capitalization).

This statement is easiest to understand if Resources are things with
persitent,
atomic identity.  What actual URIs 'identify' is sometimes like that, but not
always.  It is not that way for a much-used subclass of URIs: search URLs. 
For
search URLs there is no persistent identity, no "capital-R Resource posessing
persistent identity" which is "referred to."

Search URLs do not operate by referring to a persistent identity.  They
operate
by describing characteristics of resources which, if they exist, would be of
interest to the party originating the search URL.  They characterize demand. 
They don't identify supply.

They do identify the engine whose service they seek to enlist.  However, the
real meat of the search URL, and what is unique to, and captured in, the
URL is
not the identity of the resource or resources sought, but their
characteristics.

The binding of search URLs to resources is done on the fly by the search
engine.  This binding does not have, nor would it be good for it to have,
persistence.

There is a moral to be learned, here.

Sandro has a pretty good discussion at 
<<http://www.w3.org/2001/03/identification-problem/opacity-spectrum.html>ht
tp://www.w3.org/2001/03/identification-problem/opacity-spectrum.html> 
of how names are derived from descriptions.  In other words, names are
descriptions.  But not all descriptions are names.  There are plenty of
application requirements for reference-by-description where the reference
is by
nature _not_ a matter of identity.

That is the kicker.  It's not just some anomaly that search services are
prominent in the de_facto Web.  What we need to be able to embed in the
retained, encoded form of discourses are "references by description" in
general, and not only references by description of the special case
"references
by identity."  

Consider how one writes rules: one asserts something about referends which are
acceptor-patterns; definite as to intension, indefinite as to extension.  And
we need to be able to write rules.  So it's not just an accident that advances
in searching have powered the winning portals to the top of the Web heap. 
This
is closely aligned with the energy that fuels the Web.  We need to reflect
this
in our concepts of references and the retained representation of these
references.

'Identifiers' is a good mnemonic handle for URIs but not, if literal
construed,
a satisfactory definition.

Al  

Received on Wednesday, 2 May 2001 12:36:47 UTC