Re: [multipathtcp] [Int-area] SOCKS 6 Draft

[Joe Touch <[e-mail]>]

On 7/6/2017 1:41 AM, Dragoș Niculescu wrote:
> ----- On Jul 5, 2017, at 7:59 PM, Joe Touch [e-mail] wrote:
>
>> On 7/5/2017 9:39 AM, Vladimir Olteanu wrote:
>>
>>
>> It can also be stacked as many times as desired for arbitrarily long proxy
>> chains. However:
>> * We avoid using the SYN's payload as extra option space (which, I think, goes
>> against TCP's core philosophy).
>>
>> [Med] This is also true for MP_CONVERT Information Element which is not a TCP
>> option, but a data supplied for proxy purposes in the SYN payload.
>> Fair enough, but this is not a purely layer 5+ protocol. It seems that you are
>> strongly tied to TFO (between the client and the proxy). MP_CONVERT must be
>> part of the SYN's payload, because the following SYN+ACK depends on the
>> contents of MP_CONVERT and signals that the remote server has accepted your
>> connection.
>> The biggest impact of including non-data information in the SYN payload area is
>> that it completely defeats graceful fallback for SYN receivers that don't
>> support the option. As you note, it can be *more* safe when tied to out-of-band
>> context (e.g., prior TFO support), but TCP has NO requirement that such context
>> is absolutely maintained across different connections. You might be speaking to
>> a different stack or demuxed off to a different virtual host behind a load
>> balancer.
>>
>> Ultimately, putting any non-data info in the SYN payload violates the
>> requirement that TCP options can be ignored by receivers that don't support
>> them *without* impacting the ability of *that* connection attempt to succeed.
>>
>> Joe
> SOCKSv6 proposal makes use of extra data in the SYN (SOCKS data, and user data), but 
> its correctness and backward compatibility does not depend on TFO, only its RTT performance. 
> In fact, when TFO is not available neither between client and proxy, nor between proxy and 
> server the SOCKSv6 RTT is still lower than SOCKSv4 and SOCKSv5. But TFO is likely to be the most 
> common case in the future - Linux kernel has TFO client side on by default since 3.12 
> (November 2013)[1], and it seems to be the default in all Android phones and default 
> Linux installs.  
What happens with a legacy receiver?

Joe