= URLs prone to decay =

* Things which are dynamically generated and feed a content identifier as a parameter to some script/CMS/whatever
* Filetype or platform specific addressing: things which end in .asp, .cfm, .php, .pl, .cgi (the last two seeming slightly more stable, probably due to usage patterns)
** .htm, .html where replaced by dynamically generated content
** index.html or a nonstandard directory index instead of /
** things which live in /cgi-bin
* Detailed but essentially subjective and arbitrary hierarchies, especially when composed of abbreviations
* Hierarchies generally - the more nested, the more fragile a given resource will be
* Most meaningless content id strings
* URLs dependent on user ids, especially those which expire (any paid account, students, employees)
* anything with case-sensitivity.

= stuff =

* [http://informationr.net/ir/9-2/paper174.html A longitudinal study of Web pages continued: a consideration of document persistence] - Nice title, but it's a tiny sample.
* [http://www.mikeschinkel.com/blog/welldesignedurlsarebeautiful.aspx well designed URLs are beautiful] - unfortunately the URL is followed by an ".aspx" extension...