Here's an idea for a peer-to-peer client search engine:
you use a client side application and this application has two sides,
one to your own computer (watch the permissions) and one to your peers of
choice, whatever registry you choose to visit. You pick up these registries with
a central registry _consisting of nothing else than a pure listing by text_! These may be Wikipedia and all else (Europeana incl.).
Note: I may have shared this idea with two other people, one from Google and one from Germany from the last 12 hrs.