Re: [PANRG] Question regarding multi-link work

Hi, Jens,

On Thu, Mar 24, 2022 at 12:00 PM Jens Finkhaeuser <jens@interpeer.io> wrote:

> Hi all,
>
> I'm new to the lists (tsvwg and panrg), and have been pointed here for
> multi-path related topics. I have a question if there is interest in
> something.
>
> A couple of years ago I created a very simple tunneling protocol with
> multi-link capabilities. From that I started writing a paper, which never
> ended up getting published. I was wondering if that paper might generate
> some interest as an internet draft, in which case I'd put effort into
> converting it.
>
> The work was borne out of experiences with MP-TCP for tunneling on the one
> hand, and requirements from drone communications links on the other. In
> particular, the latter requires very fast fail- or switchover between
> links, while the tunneling application makes TCP a little less than ideal.
>
> In its concrete (prototype) implementation, the protocol is UDP-based, so
> is probably comparable to MP-UDP efforts (at first glance).
>
> However, the paper concerns itself quite deliberately with the abstract
> state machine(s) and messages. As such it does not place any requirements
> on UDP/IP addressing and could just as easily be applied to serial links,
> Ethernet or any communications channel that can be identified somehow. In
> fact, I would consider its main contribution that it is as agnostic of the
> message transport as I deemed possible. This is also why I have been
> avoiding the "path" term a little.
>
> Similarly, it is not particularly concerned with bidirectional
> communications. It takes the view that a multi-link is established if at
> least one egress and one ingress link is viable.
>
> Since multi-path concerns have been discussed here, I can see few better
> places to take this. But with how far removed the paper is from any
> protocols in development at the IETF, I still wanted to sound out interest
> before taking on the work in converting this into an Internet draft.
>

Speaking only as one participant in PANRG, I think PANRG is a fine place,
at least to start, because the multipath work in IETF (at the transport
layer) is pretty much tied to a specific protocol that's being multi-pathed
- Multipath TCP is in TCPM, Multipath QUIC is in QUIC, Multipath DCCP is in
TSVWG, and if a new home for SCTP is created, as was discussed in
TSVAREA last week, that would be a fourth home.

And, surprising literally no one, I'm interested in learning more about
your thoughts.

I didn't see a link to your work-in-progress - could you share it, whether
with me or with PANRG?

Best,

Spencer

> Thank you for your initial feedback!
>
> Jens
> _______________________________________________
> Panrg mailing list
> Panrg@irtf.org
> https://www.irtf.org/mailman/listinfo/panrg
>