Re: Multipath Milestones

[Mikkel Fahnøe Jørgensen <[e-mail]>]

I briefly skimmed the multi path proposal and one thing that comes to mind
is the limit of 256 paths.
I think there could be much more than 256 paths and that path ID’s could be
varint encoded. A single address could have multiple paths over time, key
migration could be tied to paths, although two paths may still share a key.
Path swapping could potentially be used to protect against optimistic ACK
attack. I wrote an issue on segmented packet numbers:
https://github.com/quicwg/base-drafts/issues/1105 here the objective is to
avoid the packet number encryption and to provide simpler per path loss
management, but it overlaps with multi-path. I wonder how much the current
design actually needs to change in order to support multi-path if using
segmented packet numbers. There is already PATH_CHALLENGE.


Kind Regards,
Mikkel Fahnøe Jørgensen


On 16 February 2018 at 14.11.17, Quentin De Coninck (
[e-mail]) wrote:

Hello,

Le 14/02/18 à 20:22, Ian Swett a écrit :

I think the right path forward is for those particularly interested in
multipath to experiment with QUIC extensions inside v1, once the extension
mechanism is defined.  Once some experimentation has happened, we'll be in
a better place to have an informed discussion about some of the points
Christian raised in his multipath requirements draft:
https://tools.ietf.org/html/draft-huitema-quic-mpath-req-01


I totally agree that experimenting is key. That's why we implemented the
multipath extensions that we proposed last year in
https://datatracker.ietf.org/doc/draft-deconinck-multipath-quic/

This document has evolved and we'll release a new version before the London
IETF to take into account various comments received. In parallel with that,
we are finalizing an iOS application that uses multipath QUIC to evaluate
how this works in the wild. This application will perform various
measurements to analyze the benefits of multipath. If you have ideas, feel
free to contact me offline. If there is interest, I might be able to report
some results with multipath QUIC in the wild at the London IETF.

Quentin