Eh, with the new oprhan/impl rules associated and uh... not... associated(...?) types have developed quite a schism.

Well, the first two confusingly change two variables at once: the things that are the same (which I presume is what kimundi meant) are

    fn foo<S, T: MyTrait<S>>(elem: T) { ... }

    fn foo<S, T>(elem: T) where T: MyTrait<S> { ... }

and

    fn foo<S, T: MyTrait<Thing = S>>(elem: T) { ... }

    fn foo<S, T>(elem: T) where T: MyTrait<Thing = S> { ... }