Re: [LOOPS] New Version Notification for draft-amend-iccrg-multipath-reordering-00.txt

[Carsten Bormann <[e-mail]>]

Hi Markus,

On 2020-07-10, at 10:43, <[e-mail]> <[e-mail]> wrote:
> 
> Even if Carsten already informed me, that multipath is not in scope initialy at LOOPS, I could imagine that some of the techniques in the draft are covered by your proposed charter https://github.com/loops-wg/charter.

The resequencing part is not, as LOOPS does not resequence.

In previous discussions, we have always considered resequencing as an optional service that could be provided unilaterally at the egress (we have different experiences with and thus different opinions on resequencing).
In many cases, the data that enable recovery (retransmission or reconstruction) are all that is needed to run a (non-reliable, as needed on the IP layer) resequencing process; this would use static expiration (Section 5.3 in your draft) or adaptive expiration (Section 5.4).  The adaptation mechanisms in the latter would benefit from some measurement information being available at the egress, but that is probably already needed to control the formation of return channel information (ACKs etc.) or for reconstruction (symbol numbers for FEC).
This has always been implicit; I’m not sure we want to make supplying measurement information for a resequencing process an explicit objective.

Section 6 (recover mechanisms) of course is the essence of LOOPS.  We are not discussing the multipath aspects for the LOOPS MVP we describe in the charter, but LOOPS MVP could run either per-path or on the aggregate in a multipath environment.  I think that if a protocol for this goes forward, that could be based on LOOPS even if that doesn’t cover the multipath use case fully yet.

Grüße, Carsten