Api for ed25519::keypair could benefit from a newtype

[Ryman]

pub fn keypair(seed: &[u8]) -> ([u8; 64], [u8; 32]) => pub fn keypair(seed: &[u8]) -> (Secret, Public)

let foo = keypair(..); .. foo.0 .. can be non obvious for readers.

[xorxornop]

It's more intended to be used in the destructured tuple form:

let (priv, pub) = keypair(seed);

... but yeah, a struct for PKC keypairs would probably be a good idea. Similarly with symmetric keys, actually, as keys should be made different to arrays of bytes. This is a large-ish undertaking, however, and probably something best decided on as a group or by the maintainer, as it is a question of policy: do we implement this functionality for users, taking an opinionated stance, or let them do it how they (the users) like?

[cheme]

Another notable possible misuse, the size of the seed should be enforced by its type (seed of size different than 32 bytes break the algo(only visible when using verify)). Yet I do not have a practical answer for this ([u8;32] is not really nice to use).