Architecture of theWorld WideWeb - Social Semantics: The Search for Meaning on the Web - page 17

2.2.1 Protocols

A protocol is a convention for transmitting information between two or more
agents , a broad definition that encompasses everything from computer protocols
like TCP/IP to conventions in natural language like those employed in diplomacy. A
protocol often specifies more than just the particular encoding, but also may attempt
to specify the interpretation of this encoding and the meaningful behaviour that the
sense of the information should engender in an agent. An agent is any thing capable
of interacting via a protocol . These are often called a 'user agent' on the Web, and
the term covers both web-browsers, humans, web spiders, and even combinations
such as humans operating web-browsers. A payload is the information transmitted
by a protocol . Galloway notes that protocols are “the principle of organization native
to computers in distributed networks” and that agreement on protocols are necessary
for any sort of network to succeed in the acts of communication (2004). 5 The
paradigmatic case of a protocol is TCP/IP, where the payload transmitted is just bits
in the body of the message, with the header being used by TCP to ensure the lossless
delivery of said bits. TCP/IP transmits strictly an encoding of data as bits and does
not force any particular interpretation on the bits; the payload could be a picture
of the Eiffel Tower, web-pages about the Eiffel Tower, or just meaningless random
bits. All TCP/IP does is move some particular bits from one individual computer to
another, and any language that is built on top of the bit-level are strictly outside the
bounds of TCP/IP. Since these bits are usually communication with some purpose,
the payload of the protocol is almost always an encoding on a level of abstraction
above and beyond that of the raw bits themselves.
TheWebisbasedona client-server architecture , meaning that protocols take
the form of a request for information and a response with information .The client
is defined as the agent that is requesting information and the server is defined
as the agent that is responding to the request . In a protocol, an endpoint is any
process that either requests or responds to a protocol , and so includes both client
and servers. The client is often called a user-agent since it is the user of the Web.
A user-agent may be anything from a web-browser to some sort of automated
reasoning engine that is working on behalf of another agent, often the specifically
human user. The main protocol in this exposition will be the HyperText Transfer
Protocol (HTTP), as most recently defined by IETF RFC 2616 (Fielding et al. 1999).
HTTP is a protocol originally intended for the transfer of hypertext documents,
although its now ubiquitous nature often lets it be used for the transfer of almost any
encoding over the Web, such as its use to transfer XML-based SOAP (originally the
Simple Object Access Protocol ) messages in Web Services (Box et al. 2000). HTTP
consists of sending a method , a request for a certain type of response from a user-
agent to the server , including information that may change the state of the server.

5 Although unlike Galloway, who descends into a sort of postmodern paranoia of protocols, we
recognize them as the very conditions of collectivity.