The Semantic Web - Social Semantics: The Search for Meaning on the Web - page 65

So, if any information resource is any resource which can have its content realized
as a Web representation, then information resources must be on some level digital
so that they can be encoded as Web representations. Then both the text of Moby
Dick and a web-page about the Eiffel Tower are information resources, even if
they are not currently Web-accessible. Digital information can be transmitted via
digital encodings, and so can in theory be on the Web by being realized as Web
representations, even if the resource does not allow access to Web representations
at a given time. Lastly, a particular edition of Moby Dick, or Moby Dick in French,
or even some RDF triples about Moby Dick , are all information resources, with
various encodings specified at certain levels of abstraction. It appears that the best
story we have to tell about the rather clumsy term 'non-information resource' is that
a non-information resource is a thing that is analogue and so resists direct digital
encoding, but can only be indirectly encoded via representations of the thing in a
suitable language. This would then at least be the rather odd combination of physical
entities and abstract concepts. So the Eiffel Tower itself, Tim Berners-Lee himself,
the integers, and a particular topic at a given point in space-time (i.e. on a particular
shelf!) are all non-information resources.
Should there be a class to which a web-page about the Eiffel Tower belongs
but the text of some as-of-yet unwritten novel does not? In other words, it seems
that the class of information resources is too large, and we need a term for things
that are actually accessible over the Web at a given time. We call this kind of thing a
Web re so urc e , an information resource that has accessible Web representations that
realize its information. A Web resource can then be thought of as a mapping from
time of request to a series of Web representation responses, where the information
realized by those Web representations are the Web resource. This definition is close
in spirit to the original pre-Semantic Web thinking behind resources in IETF 1630,
as well as in IETF RFC 2616 where a 'resource' is defined as “a network data
object or service ” and coherent with Engelbart's original use of the term 'resource'
(Engelbart and Ruilifson 1999; Fielding et al. 1999). A Semantic Web resource is
a resource that allows access to Semantic Web documents .
The distinction between information resources and non-information resources
has real effects. When the average hacker on the streets wants to add some informa-
tion to the Semantic Web, the first task is to mint a new URI for the resource at hand,
and the second task is to make some of this new information available as a Web
representation. However, should a Web representation be accessible from a URI for
a non-information resource? If so, would this confuse the non-information resource
itself with a Web resource that merely represents that resource. Yet how else would
fulfilling the Principle of Self-Description for Semantic Web resources be possible?
To refuse to allow access to any Web representations would make the Semantic
Web completely separate from the Web. Non-information resources need associated
descriptions , resources that have as their primary purpose the representation,
however incomplete, of some non-information resource . In other words, associated
descriptions are classical examples of metadata. According to the TAG, since the
associated description is a separate thing from the non-information resource it
represents, the non-information should be given a separate URI. This would fulfill