Ben Kaduk's Recovery Comment 2

[LPardue]

@kaduk said:

Section 5.3

After the handshake is confirmed, any acknowledgement delays reported
by the peer that are greater than the peer's max_ack_delay are
attributed to unintentional but potentially repeating delays, such as
scheduler latency at the peer or loss of previous acknowledgements.

I don't think I understand how loss of previous acknowledgements could
induce a peer to delay sending an ack longer than its max_ack_delay
after an ack-eliciting packet.

[janaiyengar]

ack_delay in an ACK frame is recorded as the time since the largest_acknowledged packet was received. If the largest_acknowledged does not change as subsequent ACK frames are sent by the receiver, ack_delay only increases in subsequent ACK frames. Now, if there's no acknowledgement loss, the receiver of acknowledgements would read the first ack_delay in the first such ACK frame. If there is loss however, the receiver of acknowledgements would read only subsequent ACK frames, which might have ack_delay that is larger than max_ack_delay.

Does that help?

I propose closing with no action.

[larseggert]

@kaduk does this explanation and proposed resolution work for you?

[kaduk]

That helps, yes, and thank you. I think I was confused about which direction of ACKs were getting lost.
"loss of previous acknowledgments of the same packet" or "loss of previous attempts to acknowledge the packet in question" would have averted my confusion, though I concede that neither flows terribly well.

[larseggert]

OK, closing with no action as agreed.