Re: [Dcrup] Hey, crypto experts, what signing algorithm should we add

[Scott Kitterman <[e-mail]>]

On Saturday, June 10, 2017 03:28:57 PM John R Levine wrote:
> > Yes, the key sizes are fixed depending on the curve.  The encodings of
> > curves are different and not self-describing.  If you see a string of
> > bytes lying on the ground, you can't necessarily tell anything about the
> > key -- if it is one, it's size, and which curve applies.  You need
> > external state such as knowing the signing algorithm; that shouldn't be
> > a problem.
> 
> That's no problem, the DKIM key record has a field for the algorithm.
> 
> > Supporting Ed25519 will not come with zero code changes.  But minimal I
> > think.
> Since we have a lot of lazy programmers who will need to do this, how
> different is ED25519 sign/verify from RSA?

For Python (I know, not everyone cares, but I do), PyNaCl offers a maintained 
and reasonably simple API for this:

https://pynacl.readthedocs.io/en/latest/signing/#example

As an implementer that would need to implement this, it seems eminently 
doable.  I'd prefer something supported in the Python standard library, but 
I'm certainly not holding my breath for it.

Scott K