Re: [jose] CFRG ECC in JOSE - thumbprint

[Ilari Liusvaara <[e-mail]>]

On Tue, Nov 10, 2015 at 08:22:53PM -0500, Justin Richer wrote:
> I would lean toward using a new “kty” value for these as the syntax
> is different from existing ones. This will help parsers and existing
> implementations add this in without adding special processing rules:
> code is already set up to branch on “kty” so let’s keep that
> behavior. Note you can still reuse parts of the key definition
> (like “d” is found in both RSA and EC keypairs) without having to
> overload a new syntax since the kty defines a new namespace,
> effectively. I suggest a value of “ED” since they’re all “edwards”
> curves from my quick read.

1) Actually, I think "okp" (Octet Key Pair) or something like that
might be more descriptive, since these are keypairs with no structure
outside the box ("oct" or whatever won't do, since that's symmetric,
not asymmetric). This also holds for X25519 and X448.

2) IIRC, the definition of "EC" talks about all defined curves having
"y", but not necressarily all curves having "y". Might be bad idea
implementation-wise to use that option however (I did encounter
comments about this pre-draft that indicated that options realistically
available are less than what JWK provodes).

3) I think it is a bad idea not to vary key fingerprint on algorithm.
However, that ship has already sailed.


-Ilari