Ever watched youself browsing e.g. a web forum?
Noticed that you viewed the same page several times?
Well, this means extraordinary and useless load for your server if there's no caching mechanism implemented in the web application. Even if there is some file or db cache you can still improve performance with implementing some http cache.
The easiest way is using PEARs HTTP::Header:
// only cache a few seconds for frequently
// changing pages like a forum;
// Header_Cache will exit automatically with
// "HTTP 304 Not Modified" if the page is
// requested twice within 3 seconds
$cache = &new HTTP_Header_Cache(3, 'seconds');
// ... load from file/db cache
// ... or rebuild page
Probably the greatest caveat of using HTTP caching is, that our beloved Internet Explorer with standard settings doesn't even send an request to the server if it just notices a "Last-Modified" header (sic!), so one'd have to press the reload button (without CTRL) that a new page is displayed after the defined amount of time. A dead end again? Decide yourself...
I finally managed to implement support for Internet Explorer in HTTP_Header 1.1.0 :D