Re: [multipathtcp] Preparing for Prague meeting - things to delete or add to MPTCP protocol bis

[Christoph Paasch <[e-mail]>]

Hello,

On 06/07/15 - 18:01:21, [e-mail] wrote:
> In the Prague meeting, we aim to move forward the protocol bis doc (to
> meeting our Cahrter objetive for a standards track version of RFC6824)
> 
> 
> 
> One topic we'd like to discuss is if there's anything that should be
> deleted from or added to the current draft,
> <https://tools.ietf.org/wg/mptcp/draft-ietf-mptcp-rfc6824bis/>
> https://tools.ietf.org/wg/mptcp/draft-ietf-mptcp-rfc6824bis/ - based on
> operational and implementation experience.
> 
> 
> 
> Any proposals?

Might we consider to include something similar to draft-paasch-mptcp-syncookies-00
in RFC6824bis, to better handle MPTCP on stateless webservers?

We received offline-comments from Olivier and Alan on the draft and still
have to update it. However, neither I nor Anumita won't be able to attend the
IETF in Prague.

draft-paasch-mptcp-syncookies-00 however also depends on whether we intend
to bump the MPTCP-version number up to 1 for RFC6824bis.

> Looking at the (expired) draft summarising the various implementations,
> 
> https://tools.ietf.org/html/draft-eardley-mptcp-implementations-survey-02
> 
> one thing that may be worth discussing is whether we need both 4 & 8 byte
> Data ack & DSS
> 
> <<   All implementations support 4 bytes "Data ACK" and "Data sequence
> number" fields, and will interoperate with an implementation sending 8
> bytes.  Implementation 1 uses only 4 bytes fields; if an implementation
> sends an 8 byte data sequence number it replies with a 4 byte data ack.
> 
> >>

I think that both, 8 bytes Data-ACK and DSS are necessary, because it may
(in theory) be possible that a segment on one subflow gets delayed while the
other one advances the sequence numbers by 2^32 bytes. In that case this old
segment will be considered as valid. 64-bit sequence numbers will protect
against this.

> I don't think the implementation survey suggests that there's anything
> else worth considering deleting - are there any other proposals?
> 
> 
> 
> As to things to add, there are a few proposals for some (minor?) additions
> in the recent set of drafts from Olivier Bonaventure and team.

+1 on draft-bonaventure-mptcp-exp-option-00, as it allows to experiment to
figure out which other signalling mechanisms would be useful to control the
scheduling across the subflows or policies for subflow creation.



Cheers,
Christoph