As another angle on the long URL theme, Jake Howlett pointed out that the URL I used for my Amazon wishlist was bogus: it always showed the current user’s wishlist, rather than mine.

What happened, of course, is I visited my wishlist, and copied the URL. It had what seemed like a long unique number at the end of it, so I figured it was mine: http://www.amazon.com/exec/obidos/wishlist/104-3115277-8167128. Actually, the random junk on the end is a session id or something (the same url with no random junk also takes you to your own wishlist: http://www.amazon.com/exec/obidos/wishlist).

Visiting Amazon again, I tried to find a good URL for others to use to get to my wishlist. I couldn’t find one: every time I got to my wishlist, Amazon knew it was me, and showed it to me in ways that wouldn’t work for others. In the end, I sent myself one of the cheezy emails Amazon would like me to send to all my friends announcing my wishlist. It had a URL that worked. (Actually, Jake sent me a URL, but I wanted to work through it myself).

All this brings up another point about long confusing URLs. Dynamic URLs can bring the power of an API to a website (in that they let you craft queries with URLs). But they can also obscure the structure of a site by bringing you to the same place with different URLs (all the different ways I could see my wishlist, for example), or to different places via the same URL (the URL that showed each person their own wishlist).

How much should web site designers make their URL structure transparent? How much can they anticipate all the ways people will want to use their sites? The permalink artifact common in weblogs is an example of providing a clear and direct mapping between content on a site and URLs that map to that content. If only all sites could be as direct.