I agree! I'd never thought of using tor, but I had very similar ideas about an engine that doesn't index a site if it has ads, etc.
I had an idea of using a delayed response in order to lessen the hardware requirements. So a user submits a query, then checks back later for a response. A little bit like a traditional library, haha!
Anyway, i'm really glad to hear others are doing similar stuff, I think there is a market. The info is out there, it's just lost in a sea of noise.
You might be interested in Ted Unangst's miniwebproxy:
Is there some more information about it? I found that it is done by https://www.wireservice.ca Is there a technical info?
I was wondering how big the index is. Because it may be just possible to serve the multi-gigabyte index with torrent. Then the update mechanism would serve only deltas. It could be then used locally. However index would probably have to be without content - so no snippets. I don't know if it would be useful that way.
It could use a comparison site. Side by side with Google for the same search. I would certainly want to use it this way for a while.
I think this is the way for the alternative search engine to somewhat succeed. Because I think it's almost impossible to compete with Google at being Google. Look at how Bing manages.