Lists Home |
Date Index |
Lyndon J B Nixon wrote:
> >no Semantic Web systems require parsing of a URI in order to draw
conclusions about the URI
> granted, but in the case of an automated system using the Semantic Web to
aggregrate retrievable resources
> about a given subject, is it not necessary to be able to determine if a
> given URI contains some retrievable resource at that location?
Retrieval returns a representation of a resource not the resource. "Semantic
Web" systems such as RDF etc, don't directly deal with representations,
rather they are used to make statements about resources (not directly to
make statements about representations of resources -- of course the
representation may itself be a resource ...).
You are asking for a "test" that is not built into RDF, so you'd need to
build this test yourself.
> possible solutions at present are:
> (1) do a HTTP GET on each URI, and see if you get a response with a MIME
> (2) in RDF, impose some standard means of expressing that the given URI
> is, in fact, a retrievable resource. In Descriptions, it could be a
> distinction between rdf:about (retrievable) and rdf:ID
> (non-retrievable), although thats certainly not part of the standard.
No, there is no such distinction between rdf:about and rdf:ID.
> could be the use of DC:identifier and DC:format to indicate an URI for a
> retrievable resource and its MIME type.
You might certainly use a predicate in *any* namespace to indicate whatever
you want, but RDF itself doesn't "understand" the meaning of your predicate,
nor can it somehow understand that you intend to make a statement about a
representation -- in RDF you are always making a statement about a resource.
> Maybe it would have been easier for this scenario if long long ago HTTP
> URIs had been restricted to retrievable resources and other formats used
> for non-retrievable resources. However this would have either
> necessitated URI parsing or in fact the same approach as now, i.e.
> determining an additional property for resources which can indicate the
> type of the resource.
Google, again, is an excellent example of a system that allows distinguishes
between resources that have (non blank) representations and resources that
don't. It works with all sorts of "http" URIs.
> I won't look good for the Semantic Web if intelligent search agents are
> correctly returning URIs relating to the search term but many of these
> URIs don't point anywhere, I suppose at least there should be some
> description (e.g. Dublin Core) applied to non-retrievable URIs so that
> users can retrieve something about search matches.
Again in RDF (as an example of a (capital "S") "Semantic Web" system), you
are directly making statements about resources and for the most part it is
entirely irrelevent whether or not they have representations (i.e. 'point
somewhere'). You can make statements about:
b) pieces of XML
c) literal strings
d) datatyped literals ... if/when RDF datatypes get sorted out
It is not the intention of, for example, an RDF query to return only
URIreferences that 'point to something' retrievable, it would return
whatever is in the system. But yes you can add a property to any resource
(make a statement about its URIreference) that could be used to filter such
an RDF query. Again, this is something you'd need to add, not something
built into the system.