Re: SSL/TLS everywhere fail

> On 7 Dec 2015, at 12:45, Poul-Henning Kamp <phk@phk.freebsd.dk> wrote:
> 
>> Yes, PSK is clearly totally acceptable. The draft should say so.
> 
> You know, I'd actually prefer the draft isn't bloated with
> boilerplate text like that.  It should concentrate on the
> task at hand and simply caution:
> 
>  "We remind the reader that Key-distribution is the only really
>  hard cryptographic problem, do not take it lightly."

Here I disagree, I simply don’t think that goes far enough. Ambiguity in RFCs is bad. We need clarity about what is and is not acceptable. At the very least we need guidance on how to evaluate the options available.

Saying “key distribution is hard, you should look up how to do it” would, in my view, represent nothing less than a dereliction of duty on our part. In my mind it’s like having a specification for GPS and saying “Boy, gravity sure messes with clocks, that’s probably important here”.

I can sympathise with not wanting a content-signing draft to get too large. Fine. Let’s take draft-thomson-signing and draft-thomson-encryption, and have them both normatively reference a draft that talks about key distribution. We don’t have to detail it in those drafts, but in my view we absolutely have to talk about it somewhere.

(NB: There may be a suitable draft on key distribution that has come out of some other working group which we could reference that I’m not aware of. That would be fine. I don’t need it to be the product of HTTPbis, but I think it’s just unacceptable for us to say “use your best judgement”.)

Cory