Re: [Cfrg] Help with the use of contexts

[Stephen Farrell <[e-mail]>]

On 16/01/17 20:09, D. J. Bernstein wrote:
> I would really like to see this unnecessary complexity eliminated from
> CFRG's signature specifications.

I'm relatively neutral on the use or non-use of contexts,
but lean more towards Dan's position that the API changes
involved mean that practically, it's better to not demand
contexts.

However, I really do wish that CFRG specs would not offer
both choices - that will simply lead to repeating this
discussion each time an IETF protocol wants to use the CFRG
spec. And of course, different decisions will be made over
time, leading to slightly more mess than would otherwise
exist. That's not a showstopper thing, but life will be
better if the choice is not offered.

So, I'd support eliminating contexts from CFRG specs and
saying something like "if you want that, and it's not a
bad idea for avoiding cross-protocol attacks, then do it
at a layer above the crypto API."

Cheers,
S.

PS: random idea - I wonder if analysis of wireshark
dissector source code, or some application calling
such code, might be a fine way to spot potential
cross-protocol attacks - anyone know if that's been
tried?