This is an excellent idea! Maybe we could use some sort of regex with the URL that would be done by the site so that the cpan preference might look like this:

cpan rule: s/::/-/g
cpan url:  http://theoryx5.uwinnipeg.ca/CPAN/data/$url
[download]

or, for those who like to get a menu for the module:

cpan rule: s/::/-/g
cpan url:  http://theoryx5.uwinnipeg.ca/mod_perl/cpan-search?dist=$url
[download]

or maybe I have a customised CPAN handler on my machine that will install the module:

cpan rule: 
cpan url:  http://127.0.0.1/cgi-bin/cpan-install.pl?$url
[download]

If this is implemented it means we can all use our favorite information sources. (We'd need to change the google command to 'websearch://searchterm' though)