Architecture of theWorld WideWeb - Social Semantics: The Search for Meaning on the Web - page 30

User Agent

Server

Fig. 2.5 A user agent accessing a resource

may cause a redirection to yet another URI and so another server, and so on possibly
multiple times, until an HTTP entity may finally be retrieved.
One of the most confusing issues of the Web is that a URI does not necessarily
retrieve a single HTTP entity, but can retrieve multiple HTTP entities. This leads to a
surprising and little-known aspect of Web architecture known as content negotiation.
Content Negotiation is a mechanism defined in a protocol that makes it possible
to respond to a request with different Web representations of the same resource
depending on the preference of the user-agent . This is because information may
have multiple encodings in different languages that all encode the same sense, and
thus the same resource which should have a singular URI. A 'representation' on
the Web is then just “an entity that is subject to content negotiation” (Fielding
et al. 1999). Historically, the term 'representation' on the Web was originally
defined in HTML as “the encoding of information for interchange” (Berners-Lee
and Connolly 1993). A later definition given by the W3C did not mention content
negotiation explicitly, defining a representation on the Web as just “data that encodes
information about resource state” (Jacobs and Walsh 2004). To descend further
into a conceptual swamp, 'representation' is one of the most confusing terms in
Web architecture, as the term 'representation' is used differently across philosophy.