Re: [quicwg/base-drafts] compensation of ack_delay is fragile against errors (#2060)

[janaiyengar <[e-mail]>]

In general, I think having an inflated RTT is better than a lower than actual RTT, since inflated RTT leads to a less aggressive sender and smaller than actual RTT leads to a more aggressive sender.

Admittedly, it's possible for there to be error when latest_rtt - (inflated) ack_delay > min_rtt.  That's the error that I had considered an accepted cost of using ack_delay.

It's fair to ask what should we do about a receiver that we consider might have a bad clock... I wouldn't be opposed to ignoring ack_delay entirely for the rest of the connection... I wonder if there are counter cases that I'm missing though where this ends up being too aggressive. In general, the receiver should be rounding *down* when reporting delay, not up.

-- 
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/2060#issuecomment-442699995