Eric, > -------- Original Message -------- > Subject: Re: [P2PSIP] proposal for reload congestion control, > fragmentation, and retransmission > From: Eric Rescorla > > [Note: please figure out how to send plaintext email. HTML is > not convenient for me] > > At Tue, 03 Mar 2009 04:30:57 -0700, > > > This framing is also going to be used with TLS, so you do need<br> > > > the length there. I'd rather have a consistent framing than<br> > > > save 3 bytes.<br> > > > > > My thought section 5.2.2 is for unreliable link (UDP/DTLS) only. > > Editorial error. It was our intent that everything be framed.
It would be helpful if a section of the draft present a complete logical view of all frames and protocol layers. > > I assume the TLS record layer is sufficient for TCP. > > TLS does not preserve write boundaries. > > > > Even if framing > > is used for TCP, wouldn't it be using rfc4571 as suggested by > > draft-ietf-mmusic-ice-tcp-07? Please clarify. > > No, that's at the wrong layer--below TLS rather than above. > > -Ekr Okay, the FramedMessage is encrypted by TLS/DTLS. Then why mention demultiplexing with STUN. I don't believe STUN is send inside the TLS/DTLS layload. According to ICE, STUN should be at the outer most layer inside the TCP/UDP frame, so there are only two protocols being multiplexed inside the TCP/UDP frame: STUN and TLS/DTLS. Please confirm. Thanks --Michael _______________________________________________ P2PSIP mailing list [email protected] https://www.ietf.org/mailman/listinfo/p2psip
