I find that this can reduce overall complexity. It makes it possible to use objects that can not be copied (such as a file descriptor wrapper) and moving can in most cases not fail. Without move semantics you'd have to use smart pointers to get similar results but with extra overhead.