Re: [aqm] I-D Action: draft-ietf-aqm-recommendation-04.txt

[Simon Barber <[e-mail]>]

I do like the idea of introducting a new 'low priority' code point, were 
the top 3 bits are 0, so that legacy equipment that makes the wrong 
interpretation (higher priority) treats the traffic as BE. There is a 
mess of different interpretations out there, and the downside would be 
legacy equipment that does interpret CS1 correctly would now treat the 
traffic as BE. I don't know enough about the prevalence of legacy 
equipment to know which might be better.

Simon

On 5/24/2015 11:30 AM, Mikael Abrahamsson wrote:
> On Sun, 24 May 2015, Simon Barber wrote:
>
>> Hi Mikael,
>>
>> I can't find reference to DSCP 000010 or 000110, where are they defined?
>
> What do you mean? I mapped the drop probability bits to BE and 
> suggested this might be used.
>
>> I know the title 'assured forwarding' seems to imply better than best 
>> effort, but I think this is a mistake for AF1 - which seems to be 
>> recommended for bulk traffic that is not latency sensitive. You can't 
>> make everything high priority! I believe AF1 according to the list of 
>> recommended applications, would be better served at less than best 
>> effort priority - so the 4 queue 1a mapping based on the top 3 bits 
>> of the TOS byte would be OK. AF2 -> lower than best effort would be 
>> wrong however.
>
> This is already impossible to do in real life, see my notes that AF1 
> and AF2 being lower priority in a default configured 4-queue TOS->.1p 
> L2 environment.
>
> My suggestions is from what might be incrementally deployable in 
> todays real life networks. I don't care about history or existing 
> documents, I care about what might actually get used Internet-wide.
>