URI retrievability (formerly: The semantics of blank nodes)

In a message dated 11/14/2002 11:04:34 AM US Mountain Standard Time, 
fmanola@mitre.org writes:
>>>
>How is an RDF application processing this to know that the URI on
>line 5 is an accessible web page but the URI on line 9 is just an identity?
>Especially when both specify http: as the protocol.  I know this is off 
topic
>but some pointers on this would be helpful.  One way I can think of is
>to add the <dc:type> property to the web page and then you can assume 
>anything without a dc:type is just an identity (not a foolproof technique 
unless
>the dublin core folks add a dc:type value for "identifier" or "name").

A utility property for "retrievable" would be pretty coarse-grained
wouldn't it?  I'd be looking for more descriptive vocabulary about what
kind of thing it was, and so on, with how to access it (if it were
accessible) being additional information.  Anyway, it seems to me the
most straightforward way to find out whether something with an http URI
is accessible is to try to access it.  I know this has been debated a
lot, but certainly in some quarters there's a lot of support behind the
idea of using http URIs even for things that are just supposed to be
identities, with the idea that you might put some descriptive
information about them there (as in the RDDL document you mentioned).  
<<<

I would actually say that this is so basic that it needs to be clear.
Also as the number of RDF documents and thus URIs as identities proliferate,
the solution to attempt to access a URI is an exponential waste of network 
bandwidth.
I think Jon's email below has the seed of the answer in the route RSS
takes.  An alternate approach would be to just add another attribute
similar to rdf:about called rdf:aboutID where rdf:about refers to a 
retrievable
URI and rdf:aboutID refers to an identity.

In a message dated 11/14/2002 11:49:04 AM US Mountain Standard Time, 
jon@spin.ie writes:
>>>
>At the level of RDF both URIs are "just an identity". I agree
>that it would
>be useful for many applications to know if you could retrieve a
>representation (note that a resource is not the same as the bytes that get
>sent down the wire - see
>http://www.w3.org/TR/webarch/#resource-interactions), however
>there may not
>be a good general way of doing this, and certainly it should happen at a
>level above RDF.

One thing that should have occurred to me when I wrote this reply is that it
is interesting that in RSS the URLs that are expected to be used for
dereferencing are sent as literals, which may or may not match URIs also
used at the RDF level.
<<<

RSS 1.0 is a good example of this distinction in that it uses both rdf:about
and a <link> element.

 - Mike
----------------------------------------------------
Michael C. Daconta
Director, Web & Technology Services
www.mcbrad.com