[rscho]

Why would it be better for error handling? If you'll be using unidirectional flow only, then the point is moot. But using clp is arguably better IMO, allowing type and range checks while allowing relational execution.