Simon Willison: Why you should be using disambiguated URLs.
It's sad that its 2007 and sites still get this wrong.

Way back in 1998 I was working on Encarta Online. At some point, the marketers came in and demanded that we add some tracking information to our search results. The first search would get an "o=1" parameter, the next "o=2" and so forth, so that we could tell how far down the search list people were actually clicking.

This after I had already spent a lot of time convincing people that the link to an article always needed to be the same. That meant cleaning up our (duplicated) URL-generating code to make sure that the query string parameters in a link to an article always came in the same order, that case mattered, that all legacy querystring parameters were stripped, etc. Why is that important? Caching!

If the characters in a URL are different, the browser won't cache the item, even if it's the "same" from the server's point of view.

When viewing an encyclopedia article you end to click around the links in an article, and there tend to be clumpy webs of articles that all link to each other. Since this is 1998 and you're on a crappy dial-up line, and maybe not even a 56k line, you want to try your best to make sure a user doesn't have to download an article twice in the same session.

The "o" parameters from the search page kill this. Not only do links between articles not have an "o" parameter, but different searches have different result orders, which yields different "o"s.

I fought, I was overruled, I gave in and stopped caring.