good point... is the idea of your proposal that you wouldn't need to write those preconditions? i.e. the invariant would be enforced by the schema shape or something?