Re: [v6ops] source address failover [PI [ULA draft revision #2 Regarding isolated networks]]

[Brian E Carpenter <[e-mail]>]

On 04/06/2014 08:24, Scott Brim wrote:
> On Tue, Jun 3, 2014 at 4:16 PM, Brian E Carpenter
> <[e-mail]> wrote:
>>> Having worked actively with both shim6 and Multipath TCP and supervised
>>> their implementation in the Linux kernel, I'm convinced that solving the
>>> problem in the transport layer is much better than in the network layer.
>>> Multipath TCP can deal with the case that you discuss and we'd be happy
>>> to perform tests with the Multipath TCP implementation in the Linux
>>> kernel (see http://www.multipath-tcp.org )
>> But, of course, the great advantage of shim6, as you know, is that it
>> applies to all transport layers. MPTCP is very elegant but only works
>> for TCP applications. So we have a bit of a dilemma in this area.
>> It's not a v6ops issue though.
> 
> And the disadvantage of shim6 is that it's gone, sorry Brian but it's
> not coming back. In the meantime, while MPTCP itself has little
> deployment, the framework in it and SCTP are strongly present in HTTP
> 2.0, RTCWEB, and web browser space in general, and there is refreshing
> work being done on transport layer improvements. If and when
> multipathing proves its usefulness (imho the jury is still out, eg.
> when you have paths with highly different quality) there will be
> several convenient paths for its deployment.

My point is that there are useful learnings in shim6 as well as in MPTCP
and SCTP. However, it will be very sad if we have to re-solve these
problems in every application suite rather than solving them once
in the stack.

    Brian