Re: [quicwg/base-drafts] Duplicates are dropped (#1624)

[Kazuho Oku <[e-mail]>]

@martinthomson 
> Some of us assumed that this was how it had to work. Mainly because it is hard to think of a good reason to allow acceptance.

I prefer using MAY rather than MUST.

The reason is because there is no need to enforce every stack to be written that way. For example, a QUIC stack can do the following:
* discard a newly received packet if it contains a PN in the ack queue
* retire PNs from ack queue after some time

This would work completely fine, even though it has the chance of processing a packet with same PN more than once.

If it is a MUST, every stack will be required to have a bit vector and a minimum threshold to detect all duplicates and also other awfully reordered packets. I understand that it is possible to implement such a mechanism. However, I simply think that it is not necessary.

> Consider: what if the same packet number was used for different content? Other than it completely breaking most AEADs, what was it that was acknowledged?

My understanding is that AEAED is broken if the sender reuses a packet number for carrying different content. I think we do not need to consider how to handle a corrupt implementation in this respect.

> I can add a reference to Section 9.3 of the TLS doc. I didn't think that it was necessary, but it's easy to do.

Thank you for the addition!

-- 
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/1624#issuecomment-409469398