True, but don't those services require schema definitions and similar setup? It seems like a lot of prototyping could be sped up by simply inferring relationships based on RESTful URLs and relying on schema-less storage.
I'm not sure. It certainly couldn't handle every use case, but maybe with some convention-style configurations (e.g. a record created via a POST to /users would hash any "password" field) you could handle a bunch of typical web app scenarios.
And if that doesn't work, perhaps the schema-less prototyping is a free tier with a more robust backend service offering for the paid tier?
