Re: [multipathtcp] SOCKS 6 Draft

[Vladimir Olteanu <[e-mail]>]

Hi Yoshi,


On 07/12/2017 10:51 AM, Yoshifumi Nishida wrote:
> Hi Vlad,
>
> Thanks for the response.
>
> On Mon, Jul 10, 2017 at 4:19 PM, Vladimir Olteanu 
> <[e-mail] <mailto:[e-mail]>> wrote:
>
>     Hi Yoshi,
>
>     As a response to question 2, the client can send some initial data
>     as part of the request. It does have to wait for a reply to send
>     any more data, though. As such, the client can get a data response
>     from the server in 1 RTT, same as when contacting the server
>     directly with TFO, if:
>      * Everyone uses TFO (otherwise, RTTs are incurred only on the
>     segments where TFO is not available).
>      * Authentication is not required, or 0-RTT authentication
>     succeeds. (Otherwise, some extra RTTs are incurred, same as above.)
>
>
>     If nobody uses TFO, but authentication is not an issue (either
>     it's not needed or done in 0 RTTs), then SOCKS 6 does no worse
>     than regular TCP (2 RTTs for a data response).
>
>     As for the use case, there are two possible ways in which it can
>     be handled:
>      * The client knows about both proxies.
>      * The client knows only about proxy1, but proxy1 is configured to
>     go via proxy2 when contacting the server.
>
>
>     Assuming the client knows about both of them, it can speak SOCKS 6
>     over SOCKS 6. The relevant messages are:
>     C->P1 request(P2, auth data 1, initial data(request(S, auth data
>     2, initial data(application data))))
>     P1->P2 request(S, auth data 2, initial data(application data))
>     P2->S application data
>
>     If proxy1 is configured to use proxy2:
>     C->P1 request(S, auth data 1, initial data(application data))
>     P1->P2 request(S, auth data 2, initial data(application data))
>     P2->S application data
>
>
> Thanks for the clarification.
> So, in both cases, if C piggy-backs all requests (authentication, 
> connect, etc if any) and application data in SYN payload with TFO, P1 
> can forward it as soon as it receives and P2 can do the same thing? 
> (We might presume no authentication or 0-rtt authentication here, though.)
>
Yes, that is the case.

Vlad