> Zero impedance mismatch I see the problem here being that too many manual optimizations need to be done when implementing a schema. You start with a logical

> Zero impedance mismatch I see the problem here being that too many manual optimizations need to be done when implementing a schema. You start with a logical schema (ERD diagram) and then implement it via a physical schema with denormalizations added for efficiency (usually because the relational model has scaling limits with number of joins, or because db migrations are too difficult, or handling unstructured data without introducing a ton of new tables). The db should do the denormalization automatically, allowing the user to interface with their logical schema directly. Another reason is we can't use SQL in the browser - we have to go through many caching layers and API layers which complicate things.