Re: [quicwg/base-drafts] Describe PMTU probing that includes source connection ID for routing … (#2402)

[MikkelFJ <[e-mail]>]

> Considering that, I am not sure if it a good idea to disable the logic to split coalesced datagram into multiple packets when dropping the Handshake key. Isn't it an additional complexity to disable the logic upon the completion of the handshake?

To an extend yes, but if you do care about performance, it can be a simplification: You need the core functionality regardless, so it is mostly function calls to logic that must exist either way. And, it allows you to be more lazy performance wise when dealing with the complexities of the handshake where crypto overhead and low packet count does not make optimisations worthwhile. This means you can shift that focus to short packets.

This gets into biblical citations, but ...

>Every QUIC packet that is coalesced into a single UDP datagram is separate and complete.

but it also says that you cannot have multiple connections, and you cannot have long headers post 1-RTT. 

Technically you can coalesce a packet "sent" before the packet being coalesced, but if you have seen a short header, you can reasonably assume that any long headers are of no interest and therefore drop processing of that datagram ASAP.

-- 
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/pull/2402#issuecomment-460543273