Re: [quicwg/base-drafts] Handling of duplicate packets (#1405)

[Martin Thomson <[e-mail]>]

We can probably crib something from DTLS.  The usual method is to maintain a bitvector for the last N packets, and to drop anything that appears to be before that vector starts.  The point about stateless reset is a good one, and one I discovered when implementing it.  It's highly likely that a stateless reset will appear to be a duplicate.

One thing we need to observe is that we won't be checking that the packet is a true duplicate, only that we've (probably) seen the (apparent) packet number before.

Another thing here is that packets that are ECN marked, but not authenticated, might be carrying a valid congestion signal, but we drop those too.  Think about the usual background radiation on the Internet that passes through a congested link on its way to you.  I don't think that we need to worry about that though.

-- 
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/1405#issuecomment-393711328