Re: Three states for a binary bit (was Re: TCP (and SCTP) sucks on high speed networks)

[Alhussein Abouzeid <hussein@ee.washington.edu>]

On Tue, 5 Dec 2000, Panos GEVROS wrote:

> 
> one of the arguments for the shift of focus from the end-points to the router 
> was that end-points are no longer trusted to be well behaved (kind and 
> adaptive) like they used to be in the past,
> one way to deal with this would be to _enforce_ certain (control) behaviors on 
> the end-points,
> that could be the triumph of the end-to-end argument, scalability, even 
> differentiated services could come for free
> 
>  |It's lazy and arrogant to presume that the network designer knows what the 
>  |users need in a resource allocation algorithm, such as managing "congestion".
>  |At most, the network can detect congestion.
> 
> it would be desirable for the network to provide information which the (well 
> behaved) end-points can use to optimise their strategies,  -there will be a 
> tradeoff here between the complexity and marginal benefit from it though, ECN 
> is such a thing,
> 
> regards,
> Panos
> 
> 

Exactly. Note this does not limit the possible ways in which the network
can act, nor does it limit the work on more refined algorithms at the
sources. Also, I definitely do not claim that ECN is the only thing
possible, but I think it is a start.

Btw, in an erra of gegabit links and (soon) GHz processors, it worries me
when people think of using a bit in the packet header as "complex":)

Hussein.