Minutes IETF106: mptcp
minutes-106-mptcp-00

Multipath TCP (MPTCP)

Meeting  : IETF106, Tuesday Nov 19, 2019, 17:10 - 18:40 (Afternoon session III)
Location : Olivia
Chairs   : Philip Eardley
           Yoshifumi Nishida
AD       : Mirja Kühlewind
URL      : http://tools.ietf.org/wg/mptcp/
Note Taker: Christoph Paasch

--------------------------------------------------------------------------

1:  WG Updates - Chairs

  * Implementation Status

   iOS 13 uses MPTCP for Music & Maps
   Linux upstreaming ongoing, driven by RedHat, Intel & Tessares
   Implementation-feedback on RFC6824bis

   Mirja:
      Could still be added to the RFC. Needs to be done timely.

   Yoshi:
      Does Apple plan to implement RFC6824bis in iOS?

   Christoph:
      Not clear yet, but ATSSS 3GPP uses bis.

--------------------------------------------------------------------------

2: Multipath TCP Extension for Robust Session Establishment - Markus Amend

--------------------------------------------------------------------------

3: Some experience of MPTCP over satellite networks - Nicolas KUHN

   Markus:
      For perf you could scale up the receive window.

   Yoshi:
      Which PEP were you using, and does it support MPTCP?

   Nicolas:
      That depends on when you are running it on top of the MPTCP Linux Kernel

   Yoshi:
      Can you share the tcpdump file so people can debug it?

   Nicolas:
      Yes, can be shared.

--------------------------------------------------------------------------

4: Multipath TCP extensions for Trust & Prediction - Qi Zhang

 * Predictable Multipath TCP extension

   Spencer:
       Traditionally, TCP does not react upon link up/down events. But this
       here is different because it is not single-path like TCP. PANRG is also
       related to this.

 * Trusted Multipath TCP

   Qi: Next steps?

   Yoshi:
      If you have some experiments and present them next time.

   ????:
      Are you interested in exploring stronger notions of trust like IPSec
      tunnels,... would that help?

   Qi:
      Security should be exposed to the application from the transport.

   Brian Trammel:
      Do we need to store TCP-port as well because a CGNAT can break this.

   Qi: Maybe not.

   ???:
      For security from applicdation layer, we need some trusted information
      from the IP layer. IPSec does not provide the trust from the
      network-layer.

   Yoshi:
      Can't this be used without SAVI? Isn't there another way to provide the
      trust?

   Qi:
      Any framework can provide the information.