Re: [quicwg/base-drafts] Make in_flight the criterion for declaring loss (#2104)

[janaiyengar <[e-mail]>]

> I think this is an excellent reason to move the text about "Generating Acknowlegements" into transport, because having it in two places increases the chances of inconsistencies.

Agreed. It's on my todo for next week.

That's about what to do with acking PADDING packets. The question I'm raising is whether loss detection is tied to being in flight or to eliciting an ACK. I'm arguing that loss detection is premised on bytes counting towards in-flight, but loss detection logic triggered by receiving an ACK (which is why we use max_ack_delay in the computations).  The bytes being in flight are a necessary condition, not sufficient.

Loss detection marks a packet as lost when no ACK is received for it. This is predicated on an ACK being sent by the receiver, delayed or not, if the packet is received, which is only true for ack-eliciting packets.

PADDING of ACK-only frames is a weird special case that causes inflight to go up without being ack-eliciting. We had agreed in Montreal that if someone did this they will need to either add a PING frame in there or handle special conditions.


-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/pull/2104#issuecomment-445109347