I’ve been saying that readers should be able to guess URL’s of pages they’re looking for. It never occured to me that URL itself should be able to guess what you’re looking for.
Mark Pilgrim pointed me to “php.net’s incredible URL-based search engine“. If the URL you request from php.net is not on the site, the 404 error page is smart enough to make a good guess about what you’re looking for. It looks at what follows the domain name (e.g. http://php.net/searchterm) to see if there is an appropriate page on the site, then for an appropriate page in the online PHP manual, and finally it uses your URL as a keyword for a search.
This approach should have a lot of great applications for publishers. It allows you to remap requests for sections (sports, business, technology), topics (raiders, ual, cisco), or administrative pages (help, contact, careers) to the appropriate pages on your site or to give your readers a page of search results that meets their needs right away.
This technique is ideal for news sites and other with lots of information in lots of different sections.
The challenge would be educating readers to take advantage of this technique, until they learn to expect it — as they should.
But the benefit should be a stronger bond between readers and any site that uses this technique.