Refactor identity methods into Identity trait
#42
Conversation
Co-authored-by: Tony Arcieri <bascule@gmail.com>
src/lib.rs
Outdated
| @@ -172,3 +167,13 @@ pub trait UncompressedEncoding: Sized { | |||
| /// the point at infinity. | |||
| fn to_uncompressed(&self) -> Self::Uncompressed; | |||
| } | |||
|
|
|||
| /// Obtain or determine the identity point. | |||
There was a problem hiding this comment.
Same issue ("point" instead of "element").
I don't know how to describe this trait, which is concerning. It's not an extension trait because it's being added as a required bound. But the constant and method it defines do not make sense in a vacuum; they need to be in the context of a group.
This makes me question the premise of the PR: what is the motivating use case for generalising over both the affine and efficient forms of a curve point, such that a shared trait is needed?
There was a problem hiding this comment.
Apologies for the delay.
The original motivation was RustCrypto/traits#1170. We currently use Default to obtain the identity element, which isn't ideal and doesn't support const.
Unfortunately this was a long time ago and I don't have all the details in my head anymore, but I assume these sort of requirements will keep popping up in elliptic-curve because of it's nature: it introduces types that are generic over many different things, so it requires traits that are equally generic to support this idea.
Maybe @tarcieri has more to say on this topic.
There was a problem hiding this comment.
I can't think of a use case for such a generalization
Co-Authored-By: str4d <str4d@users.noreply.github.com>
Introduces a new trait:
Identity.Moving all
identity()andis_identity()methods to it. This enables generalizing over different point types when the identity point is needed.This is a breaking change.
Fixes #41.