[altq 1133] Re: ECN implementation

[Kenjiro Cho <kjc@csl.sony.co.jp>]

Nicolas Christin wrote:
> > In a couple of months, I guess.  But I'm not sure if I can support ECN
> > in TCP for all the BSDs...
> 
> 	In that respect, I have a very naive question... ECN is to be handled by 
> the sender and the receiver of a TCP flow. What does a queuing scheme 
> have to do with that except for marking the CE bit?
> 
> 	Is that the struct field corresponding to this CE bit in the different 
> BSDs that causes a problem? Something related?
> 
> 	Sorry if this question sounds dumb, but end-to-end congestion control and 
> per-hop scheduling and dropping seem pretty orthogonal to me.

You are right that they are orthogonal.  But, without TCP support, the
IP level ECN is essentially useless and I can't even test my
implementation...

> P.S.: Quick auxiliary question (I could look it up by myself, but I'm 
> too lazy :-): does the RFC in question specify anything about ECN nonces?

Yes, the new ECN RFC mentions nonces briefly but the details are in
a different draft.

-Kenjiro