HTTP: The Definitive Guide

Transparent Negotiation

Transparent negotiation seeks to move the load of server-driven negotiation away from the server, while minimizing message exchanges with the client by having an intermediary proxy negotiate on behalf of the client. The proxy is assumed to have knowledge of the client’s expectations and be capable of performing the negotiations on its behalf (the proxy has received the client’s expectations in the request for content). To support transparent content negotiation, the server must be able to tell proxies what request headers the server examines to determine the best match for the client’s request. The HTTP/1.1 specification does not define any mechanisms for transparent negotiation, but it does define the Vary header. Servers send Vary headers in their responses to tell intermediaries what request headers they use for content negotiation.

Caching proxies can store different copies of documents accessed via a single URL. If servers communicate their decision-making processes to caches, the caches can negotiate with clients on behalf of the servers. Caches also are great places to transcode content, because a general-purpose transcoder deployed in a cache can transcode content from any server, not just one. Transcoding of content at a cache is illustrated in Figure 17-3 and discussed in more detail later in the chapter.

Caching and Alternates

Caching of content assumes that the content can be reused later. However, caches must employ much of the decision-making ...