I agree - such a search engine would not be as general as Google or Yahoo. But in sacrificing generality it might become better at handling certain types of requests. I am sure that given such a service the community of HN will make light work of finding out what they are :)

Also it would by hypocritical that it would block sites with ads but likely have to use ads to generate revenue.

As the owner of a search engine I can tell you search ain't cheap.

Yeah. Also, how do you computationally distinguish between a commercial and non-commercial site? Loads of grey areas.

Our Natural Language / Language Heuristics Engine can do this. We do to a small extent now, we haven't gone whole hog because I'm not sure I agree that commercial sites shouldn't rank. For things we determine to be reference searches we downrank sales pages, for things we determine to be shopping searches we uprank them.

What product are you developing the engine for?