On Wed, Jun 14, 2023 at 9:15 AM Lars Eggert <[email protected]> wrote:

> sure. My reasoning when implementing this IIRC was that before knowing
> that its TP had been received, a sender knows that the peer will have an
> unknown ack_delay_exponent and so would assume 3. And so my sender would
> use that when encoding ACK frames.
>

I think that only works if sender and receiver both agree that all ACK
frames in the Initial and Handshake spaces have an ack_delay_exponent of 3,
regardless of what is set in the transport parameters. Otherwise the
receiver will interpret ACK Delay fields inconsistently under some
sequences of packet loss and reordering.

Reply via email to