Web Caching and Replication

5.8. Expanded Object Identifiers

Because objects are identified by their URLs, caches also normally use URLs to identify cached objects. However, servers sometimes serve different content for a given URL based on the value of certain request headers. We have already encountered one example of this behavior in Section 5.6, where servers customize content based on the cookie request header. More generally, servers can use other request headers to customize content or choose between different object representations. In effect, these headers become part of the object identifier. To avoid serving wrong object representations, proxies must also use these expanded identifiers for objects they cache.

The origin server can list request headers that must ...