Re: [quicwg/base-drafts] Changing the Default QUIC ACK Policy (#3529)

[IngJohEricsson <[e-mail]>]

I am kind of donkey in the middle... 
ACK thinning in networks should be discouraged. Standardization of ACK thinning mechanisms was proposed in 3GPP RAN2, this request was turned down mainly with reference to that it can harm transport protocol performance and does not work safely. Nothing stops UE (terminal and modem) vendors from implement ACK-thinning but I certainly hope that 5G UE vendors don't attempt to thin QUIC-ACKs as this is a slippery slope.   
High ACK rates is a 5G problem mainly for high band NR (=5G at e.g 28GHz). It is mainly a coverage issue as UEs have a limited output power due to practical and regulatory requirements. At the same time one need to address that end users don't only download movies today and even VoD streaming implies uplink traffic in the form of chunk requests. And then we have an increasing amount of people doing uplink streaming or upload an infinite amount of pictures of food plates and cats :-) 
Operators and 5G vendors have to deal with this and solve the problem with deployment of 5G access in various frequency bands to get the best downlink and uplink performance.
Reduced ACK rates is  of course welcome but I don't currently see a problem that the initial ACK ratio is 1:2 , to increase to 1:? as the congestion window grows, bounded by e.g an RTT/4 rule. But here I need to admit that I at least I have not studied this in detail.
Ergo.. I am not convinced that it is necessary to handwire the QUIC spec to 1:10 is needed but I also cannot say that it should not be done.

-- 
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/issues/3529#issuecomment-605808722