Just had an idea. A URI is a name used to refer to any document on the web, and an Xpath expression is a name used to refer to any part(s) of an XML document. Why aren’t these integrated? Wouldn’t it be lovely to be able to refer to any piece of information on the web, at whatever granularity you wish? I want URIs like this:

http://x.com/y/z.html#/html/head/title - returns the title of the page
http://x.com/y/z.html#/html/head/link[@rel=stylesheet]/@href - returns the URL of the associated stylesheet.

Of course, prior art googling reveals that it’s already been done. It’s called XPointer and it works slightly differently. Hooray for independent derivation!