Re: [quicwg/base-drafts] Padding outside QUIC packet (#3333)

[Kazuho Oku <[e-mail]>]

It seems that the text regarding the amplification limit ([Section 8.1](https://quicwg.org/base-drafts/draft-ietf-quic-transport.html#section-8.1-2)) has been last changed in response to #1863 which is marked as editorial.

While the intension of the change was to clarify that the amplification limit applies at the QUIC packet level (rather than UDP datagram level), I think @nibanks might be correct in pointing out that our understanding have been the contrary (see https://github.com/quicwg/base-drafts/issues/3333#issuecomment-573730981).

Consider the case where a client is using 0-RTT. When building the first datagram consisting of an Initial and a 0-RTT packet, the only way the client can build the datagram in one-pass is by first building the Initial packet, then the 0-RTT packet due to [the ordering requirement](https://quicwg.org/base-drafts/draft-ietf-quic-transport.html#section-12.2-3). As it is hard to tell the size of the 0-RTT packet before building it, the most likely outcome is that the 0-RTT packet would be padded. However, the server might not be able to process the 0-RTT packet, when it has lost it's resumption secret.

That means that if we are to say that the amplification limit is applied at the QUIC packet layer (for the QUIC packets that are only processed successfully), the client might end up in having very little room. I do not think that is a good outcome.

-- 
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/3333#issuecomment-573980810