Re: [xmpp] dialback in DNA

[Peter Saint-Andre - &yet <[e-mail]>]

On 10/16/14, 1:08 PM, Philipp Hancke wrote:
>> I'd need to double-check XEP-0220, one of our favorite documents. ;-)
>>
>>>>     an updated version of [XEP-0220], might define a DNA prooftype for
>>>>     dialback
>>>
>>> I just tried that and failed.
>>
>> How so?
>
> I could not figure out which proof of identity I wanted.
>
> Did I want the proof that the receiving server is who it claims to be?
> With dialback, the proof seems to be "oh yeah, it is listening on the
> port DNS says"? That's pretty bad but we know it is vulnerable to DNS
> poisoning...
>
> For the sending server, the proof is slightly better, db:verify and
> merlins spell assert the sender shares a secret with the authoritative
> server.

Thanks for the clarification.

> I just need to write up the formal prooftype definition for both.
>
> It seems this is more complicated for dialback than for PKIX or
> POSH/DANE because it's not symmetric there. OTOH, we might have
> scenarios where we have PKIX in one direction and POSH in the other.

True, that's possible. These things get messy. I think one way around 
the messiness is to say that DNA applies only to a single stream in a 
single direction. Bidi is interesting but what matters for DNA is that I 
check my server or peer.

Peter

-- 
Peter Saint-Andre
https://andyet.com/