Re: Proposal: variables, templates, and Stickey Cyber Molecules

Seth Russell wrote:
> 
> Graham Klyne wrote:
> 
> > I think it is possible to achieve this effect without requiring an explicit
> > indication that the node is a "variable", although I think there is a need
> > to indicate the scope withi which the value/name bindings associated with a
> > node may be applicable.
> 
> I responded to this in my last email; but further on you raise some very
> interesting points.
> 
> > If one has a node whose URI is unknown, then I propose to just assign it a
> > unique URI.  (I find that leaving the node anonymous is unsatisfactory
> > because one may wish to establish that different occurrences within a
> > context are indeed the same thing.)
> 
> This is where I think we need to draw a distinction between internal and external.
> Most implementations assign a unique identifier to any node they create (anonymous
> or not); the question, I think, is whether it is wise to publish that unique
> identifier to the outside world as a URI.  I would weigh in on the side that it is
> not.

It seems to me there are a couple of issues here (on the other hand, I
may be misunderstanding what the "universal" in "URI" means).  One is
the implication of using the word "publish" in the quotation above.  I
may be wrong, but just because you assign a resource a URI doesn't
necessarily mean the URI gets (or ought to be)
published/widely-distributed.  It's just that the resource is uniquely
identified within a universal scope.  There are certainly lots of Web
resources whose URLs aren't widely published, or which are only
accessible from within a given context (using a loose definition of
"context").  That's a question of access control rather than unique
identity.  So I think your distinction between internal and external (we
could even have varying degrees of internalness/externalness) is a good
one, but that seems to me distinct from whether you can use URIs for
"internal" references or not.  

A second issue is that (and again I may be wrong) it's not my
understanding that the "universal" in "URI" means "universal in time" in
the sense that the same URI necessarily identifies the same resource for
all time.  Here, I'm not talking about the fact that, e.g., a Web page
might have its content modified over time, but rather about something
analogous to the dynamic assignment of IP addresses.  For purposes of
referring to things in certain interactions, it seems to me we're going
to want identifiers that we can use in RDF expressions (hence URIs, or
at least URI-like), but that are also potentially temporary (I say
"potentially" because they may turn out to last for a while).  Now,
maybe this is "URI-like" thing is only something that *looks like* a
URI, but certainly there are plenty of URIs that identify highly
transient things.  

> 
> > Then what shall we do when we later discover further information
> > establishing that this resource node, which has by now been bound to a URI,
> > is in fact the same as some other resource in some wider context, which
> > also has a (different) URI.
> 
> Well internally if you discover that they should be the same node you just :Smush
> them together and revise all the internal pointers accordingly.  If you haven't
> published a uri for either of them externally then there is no problem whatsoever.
> 
> > One answer is to declare that the two URIs are in fact bound to the same
> > resource.  But this goes against the view held by some that URI:Resource
> > mapping is 1:1-onto.
> 
> If you cannot draw a distinction between two nodes, then they should have the same
> internal identifier.  What the outside world does is pretty close to  chaos; but
> when you publish to it, and are consistent, then that helps to simplify the chaos.
> 
> > A more flexible approach is to draw on some idea of resource equivalence
> > that allows us to infer properties of one resource from the corresponding
> > properties of another resource with which it is equivalent.  This is the
> > approach I am planning to take in my work.
> 
> Hmmm....  this works internally and externally; however it does require that a
> whole slew of equivalences be published along with your statements.  From whence is
> the infrastructure for this kind of thing going to arise?

It's going to have to arise one way or another.  Certainly if we're
going to identify things in the world (as opposed to Web pages stored in
specific files in computers) using URIs, there really doesn't appear to
be any way to keep people (or computers) from using and publishing
multiple URIs that turn out to identify the same thing, unless we have a
lot of very strict regulation by registration authorities that people
don't seem to want.  

-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-8752

Received on Friday, 19 January 2001 11:34:19 UTC