Algorithm name should be EDDSA/EDDH?

[vlovich]

It seems like these names are more consistent with how ECDSA/ECDH is defined in the standard. Similarly, if Curve448 was ever added, it would seem to be a natural extension to EDDSA/EDDH without having to define a new algorithm family. On the other hand, maybe there are technical reasons to treat these as different families?

[armfazh]

See my comment

[vlovich]

Just so we're talking about the same thing, you are referring to the algorithm name provided to importKey/generateKey and not the namedCurve parameter that would be part of the relevant params dictionary? Does this mean that the spec would then define Ed25519KeyGenParams, Ed448KeyGenParams, Ed25519KeyAlgorithm, Ed448KeyAlgorithm + rinse & repeat for the X variants?

As an implementer or a user, it feels weirdly inconsistent with ECDSA/ECDH which don't use the curve name as the algorithm name, especially since NodeJS chose to go with having X25519/X448 as namedCurve attributes to the ECDH algorithm. Is this also driven by historical reasons?

FWIW, RFC8032 defines additional curve variants (Ed25519ph, Ed25519ctx, Ed448ph). Would these be restricted to changing just the namedCurve property of the dictionary or would they need to be separate algorithms?

I'd love to learn more about the historical reasons you mention & why it's important for the spec to implicitly encode this context. I don't have a good knowledge of the history here & not all readers of the spec would necessarily either.