Re: [quicwg/base-drafts] Reduce restrictions on valid RTT samples (#2568)

[Kazuho Oku <[e-mail]>]

My +1 as well.

I think it's reasonable to recommend endpoints to send PINGs, if it wants to calculate RTT when it is not sending data.

First of all, this is _not_ a departure from the design we have in TCP; in TCP, an endpoint that is not sending data has no way of inferring RTT from ACKs.

Second, requiring the peer to send ACKs for ACK-only packets promptly makes the specification complex. Because such a mechanism cannot be solely based on a timer (because ACK-of-ACK cannot be sent without piggybacking), even though it needs to be based on a timer (otherwise the endpoint cannot receive an ACK in a timely manner).

Considering these two aspects, I think using PINGs is simpler. Also, endpoints can choose different strategies on how frequent they would send PING; some might send a PING every one second (or 1 RTT), or a download-only application might not send PINGs at all.

-- 
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/2568#issuecomment-479683624