Re: [multipathtcp] High-level design decisions /architecture

[Costin Raiciu <[e-mail]>]

Hi,

>> I don't see that we do. The general principle is that an endpoint  
>> should
>> use the ports that are currently in use at the other endpoint on  
>> its new
>> connections, to maximise getting through firewalls and  
>> middleboxes, and
>> also to support any port-based QoS traffic engineering, etc. That
>> doesn't preclude an endpoint from trying other port numbers with the
>> same token, but the chances of that working would seem slim. I'm not
>> entirely sure whether we want to explicitly permit or explicitly  
>> forbid
>> that, yet.
>
> OK. Thanks for the clarification. If using different port is very  
> limited use
> I guess it's not a major thing.
I agree with Alan. There is no point in sending port numbers for  
advertised addresses, since on subsequent subflow
handshakes - which must contain the JOIN option - local demuxing is  
not done based on port, but based on the token in the JOIN option.

> Hmm.. Suppose if you have a PC with two NICs and want to use one  
> NIC only for back-up.
> When you want to download something from a server, how do you  
> specify the above requirment?
> All I can think is signaling mechanism.. Can we leave this as an  
> open issues?
This is an open issue indeed. If the endpoint is sending, it can  
control how much data to put on each subflow. When it's receiving,
our current thinking is to use ECN (or worse, packet drops) to reduce  
the bandwidth on a given subflow.

Costin