Scalar::random shouldn't require std anymore about curve25519-dalek HOT 5 CLOSED

While we're fixing up Scalar::random, maybe we should get rid of it entirely? I think there's a trait from the rand crate that types can implement in order to be "randomly generable" by any RNG.

from curve25519-dalek.

Looking into it a bit, I think that what that would look like is implementing a Distribution https://docs.rs/rand/0.5.4/rand/distributions/trait.Distribution.html for Scalars, which people would call sample on. There's also a provided sample_iter, which could be more convenient for generating sequences than what we're doing now.

from curve25519-dalek.

I'm not sure exactly what that would look like? impl rand::distributions::Distribution<Scalar> for Uniform? But Uniform is their type, so the orphan rules will hate us, so we'd need to create our own Uniform type.. it sounds slightly messy. That should probably be a new issue in any case.

from curve25519-dalek.

This is fixed in #167.

from curve25519-dalek.

We'd define our own distribution type and implement the trait for it.

from curve25519-dalek.