Architecture of theWorld WideWeb - Social Semantics: The Search for Meaning on the Web - page 47

document , a Web representation that provides more information about the
dialect . In a Web representation using this dialect, a namespace declaration
then specifies the namespace URI . In this case, the user-agent may follow these
namespace declarations in order to get the extra information needed to interpret
the Web representation. As a single Web representation may be encoded in
multiple languages, it may have multiple namespace URIs to follow.
4. Follow any links: The user-agent can follow any links. There are some links in
particular languages that may be preferred, such as the ending resource of a link
header in HTML or RDF Schema links such as rdfs:isDefinedBy links, or links
like OWL by the owl:imports . If links are typed in some fashion, each language
may define or recommend links that have the normative status, and normative
links should be preferred. However, for many kinds of links, their normative
status is unclear, so the user-agent may have to follow any sort of link as a last
resort.

Using this algorithm, the user-agent can begin searching for some information
that allows it to interpret the Web representation. It can follow the first three
guidelines and then follow the fourth, applying the above guidelines recursively.
Eventually, this recursive search should bottom out either in a program that allows
an interpretation of the Web representation (such as a rendering of a web-page or
inferences given by a Semantic Web language) or specifications given by the IETF
in plain, human-readable text, the natural bottoming point of self-description. This
final fact brings up the point that the information that gets one an interpretation is
not necessarily a program, but could be a human-readable specification that requires
a human to make the mapping from the names to the intended sense.

2.3.4 The Open World Principle

The Open World Principle states that the number of resources on the Web can
always increase . There can always be new acts of identification, carving out a new
resource from the world and identifying it with a URI. At any given moment, a new
web-page may appear on the Web, and it may or may not be linked to. This is a
consequence of the relatively decentralized creation of URIs for resources given by
the Principle of Universality and the decentralized creation of links by the Principle
of Linking. Without any centralized link index, there is no central repository of
the state of the entire Web. While approximations of the state of the entire Web
are created by indexing and caching web-pages by search engines like Google, due
to the Open World Principle, none of these alternatives will necessarily ever be
guaranteed to be complete. Imagine a web-spider updating a search engine index.
At any given moment, a new resource could be added to the Web that the web-spider
may not have crawled. So to assume that any collection of resources of the Web can
be a complete picture of the whole Web is at best impudent.