[
Lists Home |
Date Index |
Thread Index
]
[John Cowan]
> Thomas B. Passin scripsit:
>
> > I think that much confusion has arisen when non-retrievable URIs got
into
> > the mix. On top of this, RDF uses the term "Resource" in a specialized
way,
> > so that it does not really mean the same thing as a "resource" in the
first
> > story. When a URI is non-retrievable and is used to "identify"
something
> > non-retrievable - it may be the Yosemite valley or some intangible
concept
> > or whatever, then there is no act of emitting a representation that is
ever
> > going to happen.
>
> Ah, but there is. There is no *intrinsic* difficulty in
> saying that what you get when you perform a GET on the URI
> http://www.ibiblio.org/wm/paint/auth/vinci/joconde/joconde.jpg ,
> for example, is a representation not merely of a particular document,
> but of a particular painting, and for that matter a particular woman.
Right, and this amounts to story 2. No intrinsic difficulty, but there is a
real conceptual difference. In story 1, the server emits a representation
of something that _it_ has (or can assemble from other information when it
wants) - it is the _server's_ resource. In story 2, the server would have
to emit a representation of something that it does _not_ have. It is _not_
the server's resource.
In story 1, any linkage between the server's resource and some possible
original, like that particular painting, is some kind of meta thing, falling
(at least for me) into a very different category. It is like Shannon's work
on the transmission of information of information on noisy channels. The
channel (and Shannon's results) knows nothing about the nature of the
message it is transmitting, but Shannon's work applies anyway. The actual
meaning of the message is one of those "meta" things too.
> Neither the model nor the painting are on the network, of course. These
> reps of reps and reps of reps of reps are where the difficulties arise.
>
> The question then is: when we use that URI as the subject (WLG) of an
> RDF triple, which of the three are we predicating about? The document,
> the painting, or the woman? RDF makes it impossible to say.
>
Well, that is the thing. IMO, we should not use a real network-retrieving
URI as an identifier for any RDF node, because the node is not a
network-retrievable thing. A copy of an RDF _graph_ may be retrievable, but
that is a different thing. If we want to have a node represent some actual
working network resource, use some appropriate property so say so.
> Topic maps at least sharply distinguish between assertions about documents
> and assertions about the subject matter of documents, though it remains
> outside the system whether a JPEG such as this has the painting or the
> woman as its subject matter.
>
Yes, possibly, depending on just how you capture that URL into the topic
map.
Cheers,
Tom P
|
