Re: [BEHAVE] TURN Issue: Preserving bits in the IP header

[Lars Eggert <[e-mail]>]

On 2008-1-3, at 9:49, ext Rémi Denis-Courmont wrote:
> I don't know what ECN will do for UDP, and I suppose that would not  
> work for
> TCP (since it works in pair with CWR).

ECN is orthogonal to what transport is used. All an ECN mark is is a  
signal from a router on the path that says "I almost dropped this  
packet."

Various transports are free to define various appropriate responses to  
that signal. TCP decided to treat the ECN signal as if there had been  
a loss, but that is only one possibility. For example, a UDP sender  
that was using TFRC to compute a safe rate might decide to reduce its  
send rate when an ECN mark is received. Or a CBR UDP sender might  
decide to stop transmission after receiving ECN marks at a certain  
rate, in order to take load off the network a bit quicker when there  
is congestion.

So ECN marks do need to be preserved, in order to let any intended  
response mechanism kick in.

(Also see draft-briscoe-tsvwg-ecn-tunnel on ECN interactions with  
tunneling.)

Lars

_______________________________________________
Behave mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/behave