Perfortmance of paths with asymetric processing/capacity in draft-ietf-quic-ack-frequency-02

[Gorry Fairhurst <[e-mail]>]

I have more 2 thoughts about that third use-case in 
draft-ietf-quic-ack-frequency-02:

1. I  have a request to reorientate the scenario for the 3rd case in 
section 2 of draft-ietf-quic-ack-frequency

Whilst  it is true that severe capacity asymmetry for a path can result 
in ack congestion, this seems to me to be a very specific case of where 
a sub-IP network benefits from a lower volume of QUIC ACK traffic. The 
QUIC WG applicability RFC section 7, has a more encompassing case - 
which considers the impact of ACKs, rather than the current text which 
only suggests a highly asymmetric link rate. Can we can also include 
something based on the sentence there which says:

“Acknowledgments also incur forwarding costs and contribute to link 
utilization, which can impact performance over some types of network.”

2. We looked carefully at the ACK implications of various quic 
implementations, and have just completed a major research study on QUIC 
AcKs which is now published  in a paper here:

https://onlinelibrary.wiley.com/doi/full/10.1002/sat.1466

This paper is entitled "Reducing the Acknowledgement Frequency in IETF 
QUIC", and has actual measurements and analysis for various things that 
are presented  in draft-ietf-quic-ack-frequency. The conclusions of that 
work support the proposal for the ACK_Frequency frame, and hopefully 
will be interesting to the group Also, I think it would be useful to 
cite this as an informative reference example in 
draft-ietf-quic-ack-frequency!

Best wishes,

Gorry