| Thanks David Blevins, Nice effort, you are stubborn  A few concerns to continue the discussion in decreasing order:
- the new method on ConstraintValidator breaks existing implementations so we would need at least a subclass
- the programmatic API makes it harder to find where and what constraint implementations are used. This goes a bit at odds with the general declarative logic of BV
- we can't return a single CV per annotation, in the spec there can be many; one per actual validated type is the only constraint
- the down casting in the register method sucks
|