This idea is, too. (Even though it's probably been implemented already, assuming I haven't overlooked some obvious problem with the notion.)
Here it is: Seven Degrees Of Net-Peration. (See? It even comes with a suitably stupid name, so that it may be easily berated.) Simply put: You insert two web page URLs. SDONP (it's even got a bad acronym, oh joy) takes the two, and through the magic of some smart programming, shows you precisely how many links it takes to get from one of the pages to the other.
Sure, it's not the brightest idea I ever had, but the absolute barebones implementation of it might be within my grasp, it might be fun to code (oh, I hope that I will code fun things again someday) and it might prove a more useful tool than it looks. If possible, it would randomly pick links, but I doubt that's a good idea; opens it up to never-ending web-surfing. Better is a large list of domains and the domains they're deeply rooted in, to provide better odds; even that, though, provides a very real chance of infinite cycling.
Stupidly enough, it occurs to me that it would be easy enough to just create a page that linked the two, and then triumphantly state that you could get from page A to page B via the newly-generated page C. Some people might feel that was cheating, though.
