Re: [quicwg/base-drafts] Simplify the client's PTO code by allowing the server to send a PING (#3161)

[Jana Iyengar <[e-mail]>]

I have a couple of thoughts. First, this change is not necessarily an overall simplification. It makes things slightly simpler at the client, but it adds about as much code to the server.  Additional code has to be added at the server to the PTO response to only send a PING and not a retransmission when the path is not yet validated.

Second, I think it does open up an interesting attack vector that is currently absent. @kazuho notes
that there is amplification, but I don't think byte amplification is the interesting one here, since the total load is going to be < 1 MTU. The interesting amplification here is packet amplification, which is a useful attack since UDP and packet processing costs at endpoints / clients is a relatively high cost. And you can get a fair number of packets back from the server for spoofing a couple of packets, in spite of the exponential backoff.

I'll grant that the attack is a bit of a stretch, and does require that the attacker guess the server's CID and packet number. Though, as @kazuho pointed out to me offline, we allow the server to use a zero-length CID. 

I don't think this change is necessary. And on balance, I don't think this change is a good one.

-- 
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/3161#issuecomment-547700025