What I am suggesting is a P2P self managing cache which isn't going to mangle Internet packets.
Having single points of failure is bad design in my view.