[Iccrg] Incentives for ECN

[Lachlan Andrew]

Greetings Dirceu,

On 16 July 2011 04:13, Dirceu Cavendish <dirceu_cavendish at yahoo.com> wrote:
> I know of base stations that drop IP packets in advance as a way to quench
> Internet side incoming (download) traffic
> This type of use case could be an incentive for operators to use ECN,
> without throwing away application data.

Yes, that is the incentive mentioned in the draft.  However, unless
loss rates are unacceptably high (>>1%, which causes CWND < 10
packets), then the amount of application data that gets lost is not
enough of an incentive.  (If it were, then ECN would already be widely
deployed.)

> And BTW, imho, ECNs being treated as loss at end-points may present a
> dis-incentive for end-points to react to ECNs at all, since the CA designer
> may find cwnd exponential decrease too drastic a measure to be taken when
> receiving ECNs.

I'm not sure that I understand this.  If ECNs are treated as loss at
end-points, then they are reacted to.  Do you mean that you think it
would be sensible to allow a CA designer to specify a milder form of
back-off when ECN is received?  If so, I entirely agree.

Cheers,
Lachlan

-- 
Lachlan Andrew  Centre for Advanced Internet Architectures (CAIA)
Swinburne University of Technology, Melbourne, Australia
<http://caia.swin.edu.au/cv/landrew>
Ph +61 3 9214 4837