The problem is that encapsulation makes it hard to specify reasonable constraints. For example, a "getter" might come with a constraint that "x = myobj.GetX();

The problem is that encapsulation makes it hard to specify reasonable constraints. For example, a "getter" might come with a constraint that "x = myobj.GetX(); myobj.SetX(x);" should not change the observable behavior of myobj. Or that "myobj2.ConstructFromXY(myobj.GetX(), myobj.GetY());" should result in a "myobj2" that's in some ways indistinguishable from "myobj". With a "declarative" approach, these constraints are implied in the system itself.