netdev
[Top] [All Lists]

Re: RFC iptables target for selectively removing ECN

To: Harald Welte <laforge@xxxxxxxxxxxx>
Subject: Re: RFC iptables target for selectively removing ECN
From: Andreas Ferber <aferber@xxxxxxxxxxxxxxxxxxxxxxxx>
Date: Tue, 26 Feb 2002 11:44:56 +0100
Cc: Sebastian <sebastian+list02@xxxxxxxxxxxxxx>, netdev@xxxxxxxxxxx
In-reply-to: <20020226101235.G23307@xxxxxxxxxxxxxxxxxxxxxxx>; from laforge@xxxxxxxxxxxx on Tue, Feb 26, 2002 at 10:12:35AM +0100
Organization: dev/consulting GmbH
References: <20020225144047.Z23307@xxxxxxxxxxxxxxxxxxxxxxx> <20020225224721.020ccfe4.sebastian+list02@xxxxxxxxxxxxxx> <20020226101235.G23307@xxxxxxxxxxxxxxxxxxxxxxx>
Sender: owner-netdev@xxxxxxxxxxx
User-agent: Mutt/1.2.5.1i
On Tue, Feb 26, 2002 at 10:12:35AM +0100, Harald Welte wrote:
> 
> I don't need to clear the ECT codepoint in the IP header as well? Is it valid
> to receive an IP packet which has an ECT codepoint set in the IP header, but 
> no
> ECE/CWR bits in the TCP headee?

Yes. The ECN IP header bits are set by intermediate routers which are
not required to examine the TCP header to tell if ECN should be used
for this flow (e.g. in load-balancing or failover situations, they
might not even see the SYN packet, so they have absolutely no way to
tell if the connection endpoints are ECN capable). Indeed such a
situation will be quite normal once ECN is widely deployed on internet
core routers.

Andreas
-- 
       Andreas Ferber - dev/consulting GmbH - Bielefeld, FRG
     ---------------------------------------------------------
         +49 521 1365800 - af@xxxxxxxxxx - www.devcon.net

<Prev in Thread] Current Thread [Next in Thread>