[quicwg/base-drafts] MSS Clamping Support (#3330)

[Matt Corallo <[e-mail]>]

Forgive me if I didn't manage to dig up some relevant spec portion, I'm new around here :).

>From what I can find, it looks like there is no way to do any kind of simple MSS clamping as a middlebox in quic. PMTU is great in theory, but, sadly, rarely works in practice (at least if you have to support lots of potential endpoints) - ICMP failure indication is just unreliable (yay poor "security" advice), and while probing with timeouts is great, there is a significant delay associated with doing so, wasting header overhead or time waiting for MTU discovery. Even worse, sometimes links (eg modern VPNs) can have a high MTU (eg 1500) but actually *using* that MTU has worse performance than not as it creates fragmented encrypted frames. Thus, in today's internet, most routers (including iptables, Cisco iOS and JunOS) have first-class support for clamping the MSS to some lower value, which doesn't have any of the associated pitfalls. Is there any way to allow an on-path device to provide MSS information or does this fall into the "middleboxes must not edit anything about the packet flow, only ECN is allowed to provide informtion to endpoints" category?

-- 
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/3330