I'm not sure I see the analogy. Tuples are not anonymous in the same sense as I'm talking about, since you can refer to them structurally. If we were going to support shape-generic programming over structs, I think we would do so in terms of things that look like protocols, and could fit in the existing constraint model.