I'm with duckyd on this one. The types of things I keep in config files are:

• Path to useful files that Perl might need to load.
• base URL to the CSS / images / javascript and other files I might need to refer to in the HTML
• Info on how to generate a URL on the machine (not all of my stuff stays in a 'cgi-bin' directory, and I might need port info, and the name of the server)

Basically, it's all machine-specific info. Stuff like REMOTE_ADDR and the like get loaded from %ENV on a per-hit basis. You can use modules for it, but you want to call an instance of the module once per hit -- you don't want the stuff stored as globals in the module, or you're just asking for trouble. (if you have a long-running process, process B may kick off before process A finishes ... you use globals, and you don't know what A might get when it asks the module for info)

PS -- CGI can return environmental variables used in CGI processing.