Re: [Tcpcrypt] v3 of the charter

[Joe Touch <[e-mail]>]

On 4/30/2014 2:06 PM, Daniel Kahn Gillmor wrote:
> On 04/30/2014 04:17 PM, Joe Touch wrote:
>> On 4/30/2014 1:06 PM, Daniel Kahn Gillmor wrote:
>>> TCP is bidirectional, but the handshake is asymmetric: one party
>>> listen()s and accept()s, and the other party connect()s.
>>
>> TCP supports simultaneous open too. In which case, who's the client and
>> who's the server?
>
> In the simultaneous open case (does anyone actually use that in
> practice?), both parties already know something about their peers (at
> least, they know who they want to connect to).  From the perspective of
> fingerprinting concerns, that suggests that they should both be treated
> as servers, not as clients.

They're also both clients because they issue the initial (non-ACK) SYN. 
So they're both - which is why most TCP protocol specification docs tend 
not to use the term client/server.

Joe