Re: Valid representations, canonical representations, and what the SW needs from the Web...

Roy, you seem to be saying that in the REST model, one cannot use a
URI to identify the Sun.  In REST, as I read your explanation, the Sun
is not a resource.  To remove an issue about intent and errors, let
revise my two-pages-about-the-Sun example to say sun-c is always a
picture of the sun and sun-d is always text about it.  You say the
resource identified by each URI is the intended persistent sameness
associated with each URI.  If the URI authority says the resource is
the Sun itself, that authority is speaking nonsense (or metaphor).
The authority could perhaps rephrase: the resource is a web page about
the Sun (and perhaps be more specific).

Is that right?   (It seems like an excellent model to me.  I like the
term "persistent sameness".)

So if RDF folks want to identify the Sun (so they can make RDF
assertions about it), and they use a URI which they say identies the
Sun, are they violating the architectural principles of the web?
Should they do something different like say the URI is a
subjectIndicator for the sun, instead of an identifier for it?

   -- sandro