Well, how do you chose which implementation will get picked? In my experience, explicit wrappers is the cleanest way ― but then again, I also find explicit casts/conversions (which essentially are the same thing) to be the cleanest way too.
If it would be ambiguous then you would have to specify yes. But for the cases where the required type means there's only one implementation that could work, I don't see the value in having to specify.
Uh, how would such an interface look? The one that can be implemented in two ways, and yet "the required type means there's only one implementation that could work"?
I'm mainly thinking of parameterized interfaces, like `Comparable` or some kind of "Keyed<T>". Another example would be datastructures that can be iterated in multiple ways, e.g. a multidimensional array that can implement both `Iterable<Int>` and `Iterable<Row<Int>>` or something like that. You could have cases where you're passing that to a function that expects Iterable<Int> so it's obvious which implementation you want, and cases where you would have to disambiguate.
