fnv1a looks interesting. I hadn't heard of this hash but I see how it'd be extremely fast and also easy to implement.
I hadn't really thought through the merkle thing, I just kinda spat out words. I'll try not to do that in the future :-P As far as how paranoid to be... we obviously don't want to do a full table scan on writes, but could we afford a full table scan at startup to verify that the view index hasn't been corrupted since couch was last run? -Randall
