I don't understand this perspective... I was confused at first, but I read about the first third of the couchDB book and then got it, and now I am excited that

I don't understand this perspective... I was confused at first, but I read about the first third of the couchDB book and then got it, and now I am excited that a "distributed map reduce system in erlang" (or javascript) is a much better solution- essentially my query results are pre-computed by mapreduce and this should make performance much better. This is a lot easier than trying to figure out how to make SQL queries that will be performant or optimizing them later... in a previous startup we lost about 18 months solving this very problem. In fact, I see map reduce as a competitive advantage. Maybe I'm a fool... I hope not.