E.g. given a trait HasBytes with an associated field bytes: [u8], I could imagine impls of it for particular structs mapping it onto any of:
(a) a fixed-size array my_field: [u8; 666],
(b) a single my_field: u8,
(c) an empty array[], for example if the struct doesn't even contain any u8s,
(d) a dynamically-sized array my_field: [u8].
and indeed, this all makes sense. But it might be something to defer for a follow-up RFC or for the implementation.
@glaebhoerl points out that it might make sense to allow fields of "DST" type. They give some examples:
and indeed, this all makes sense. But it might be something to defer for a follow-up RFC or for the implementation.