Re: [re-ECN] Viability issue #1

[Leslie Daigle <[e-mail]>]

Hi,



John Leslie wrote:
> Leslie Daigle <[e-mail]> wrote:
>> 1/ This is about network flows, not applications or services, so it will 
>>    serve future network uses (application types).
> 
>    The idea is right -- but it's hard to comprehend.
> 
>    Rather than "about", I'd say CONEX distinguishes "congestion" within
> the forwarding networks from application or services congestion.
> 

I think this highlights that there are 2 different points to get out.

The original point I was aiming for was -- taking application-specific 
approaches to handling congestion makes it harder to handle new 
applications -- unrecognized.

I understand your point to be slightly different - that this is handling 
network congestion, as opposed to managing application and service 
behaviour.


Leslie.

>    We probably also need to introduce the point that at every forwarding
> step, there's an indication of how much "congestion" is expected from
> the flow of packets heading toward that path (one RTT ago).
> 
>    And rather than some vague "future network uses", we should perhaps
> mention delay-sensitive traffic (which now isn't served at all well).
> 
>    (I wonder if we should start another thread on what "congestion"
> means: granted, CONEX is pretty agnostic on the actual meaning, but it
> feels to me as if folks need to escape the "congestion==packet-drop"
> mindset in order to understand how CONEX can help without giving
> "unfair" advantage to marked traffic.)
> 
> --
> John Leslie <[e-mail]>
> 

-- 

-------------------------------------------------------------------
"Reality:
      Yours to discover."
                                 -- ThinkingCat
Leslie Daigle
[e-mail]
-------------------------------------------------------------------