Re: [Int-area] Middleboxes to aid the deployment of MPTCP

[Joe Touch <[e-mail]>]

On 7/19/2017 12:51 PM, Olivier Bonaventure wrote:
> Joe,
>>>>
>>>> On 7/19/2017 12:41 AM, Olivier Bonaventure wrote:
>>>>>> - IMO, TCP always needs to be able to fall back (which should be
>>>>>> true now)
>>>>>
>>>>> This is not a concern with the proposed design
>>>> Prove that is true if/when TCP-AO is enabled.
>>>
>>> I don't think that TCP-AO is a use case for the proposed converters.
>>
>> You don't get to decide that. If you use TCP, then TCP-AO could be
>> enabled on the client.
>
> The converter is not intended to be used for all TCP connections. In
> the draft we explain how an MPTCP endpoint can bypass the converter if
> the destination server supports MPTCP. For TCP-AO, my recommendation
> would be that the default policy of the client would be to never use
> the converter if TCP-AO is requested by the application.

How do you know you're using the converter? Is the initial connection to
that converter? Or does the converter hijack (the latter is the
implication of the text, AFAICT).

Joe