My yesterday's entry in Mental Arc about the difficulty of finding a website among a growing number of TLDs prompted me to write a pre-draft of an RFC for a generalized method of autodiscovery of information about and services provided by a website. The core idea is to have an "autodisc" folder at the root of a webserver containing a number of files to do things like what is done with robots.txt and favicon.ico today. In fact, robots.txt is invited to move into that directory, and a replacement is specified for favicon.ico, which would allow other formats besides ".ico". Also specified are a file for providing a one-line site description, and list of important RSS and Atom feeds on the site, and a method of delegating autodiscovery authority for portions of a website's address space. Comments are welcome.