Re: [quicwg/base-drafts] Change max_packet_size to max_datagram_size (#3471)

[Martin Thomson <[e-mail]>]

This was never intended to be a limit on datagram size.  That is, it was intended to be a limit on the length of ciphertext passed to the AEAD.

That said, this is probably the right approach.  The size of the buffers you need to catch a UDP packet are probably more relevant than the size of the AEAD inputs.

I do note, however, that the construction of the Retry packet is such that the inputs to the AEAD might be greater in length than the packet itself.  And Retry has to be sent without even looking at the client's limits.  Maybe we need to say that Retry needs to be no more than 1179 bytes in length (to allow for the largest possible connection ID to be added).  That shouldn't be a problem in practice, as there is strong motivation for the Retry token to be much shorter than that - to the point that an 1179-byte Retry probably will have a token that doesn't allow for any data in an Initial packet - but we should be clearer about our limits.

Please don't claim that this is an editorial change.  It is not.

-- 
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/3471#issuecomment-589491305