Re: [LOOPS] BOF co-chairs thinking on LOOPS next steps

Not if the egress doesn't re-sequence.

Then the worst thing that can happen (except for congestion in the LOOPS segment itself, which the LOOPS nodes should avoid) is that the retransmission doesn’t help.
It does help, though:
- for reasonably re-ordering robust TCP implementations (I’m not even talking about RACK: spurious loss recovery mechanisms such as F-RTO are also widely deployed)
- for, e.g., UDP-based applications that might not care about re-ordering  (I’m not an expert enough on RTP to know how many RTP applications could handle some degree of re-ordering too, but I suspect it’s a fair amount)
- for application-limited scenarios where a new packet might not even appear in time to create re-ordering; the last data packet of a TCP connection (tail loss) is such a case.

So I’m an advocate of “never create extra delay”. Delay is bad, and it’s hard to know what “just the right amount” of extra delay is (it depends on too many factors IMO).

Cheers,
Michael

> On Jul 25, 2019, at 3:13 PM, Marie-Jose Montpetit <marie@mjmontpetit.com> wrote:
> 
> Which begs the question if the retransmission delay will kill the application.
> 
> mjm
> 
> Marie-José Montpetit, Ph.D.
> Research Affiliate, MIT Media Laboratory
> mariejose@mjmontpetit.com <mailto:mariejose@mjmontpetit.com>
> mariejo@mit.edu <mailto:mariejo@mit.edu>
> On July 25, 2019 at 3:10:39 PM, Carsten Bormann (cabo@tzi.org <mailto:cabo@tzi.org>) wrote:
> 
>> On Jul 25, 2019, at 14:59, Spencer Dawkins at IETF <spencerdawkins.ietf@gmail.com <mailto:spencerdawkins.ietf@gmail.com>> wrote: 
>> >  
>> > […] If the LOOPS community can’t live without host-to-node, the community should add it. If the community can make use of LOOPS without host-to-node and add it later, the community might consider whether that reduces the scope of the initial proposed charter enough to help Magnus approve it :-)  
>> 
>> Right. One other place where we can structure the milestones further is by starting with retransmission mode only and add FEC mode in a second step. 
>> 
>> So this would give us a nice and small first deliverable (retransmission mode for node-to-node) that we can then recharter from. If we do this right, this doesn’t even need to increase the time up to when we deliver the whole protocol including both host-to-node and FEC. 
>> 
>> Grüße, Carsten 
>> 
> -- 
> LOOPS mailing list
> LOOPS@ietf.org
> https://www.ietf.org/mailman/listinfo/loops