Re: [quicwg/base-drafts] Path Challenge Padding and Amplification Protection (#4257)

[Jana Iyengar <[e-mail]>]

This is a response to the the discussion in #4264 on the potential performance cost of applying this limit on quiescence.

This roundtrip isn't the only perf issue on NAT rebinding:
- a server could reset the congestion controller
- the server infrastructure could send a GOAWAY since incoming packets are having to be redirected between servers, forcing a new connection for subsequent requests.

Despite the fact that we want to have continuity across NAT rebinding, we did not design for it to be seamless in terms of performance. I don't think we should start now.

Let's remember that NAT rebinding is the _uncommon case_, and typically occurs after quiescence of 30 or 60 seconds. A client app that wants to improve performance because this quiescence is a common pattern for it can always use one of many methods, including sending PINGs.

I'd strongly argue against making any recommendations. This is a question where I believe more experience is necessary to learn best practices.

-- 
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/4257#issuecomment-721457465