W3C home > Mailing lists > Public > public-webrtc@w3.org > November 2017

Re: A very short extension spec: DSCP codepoint control

From: Harald Alvestrand <harald@alvestrand.no>
Date: Wed, 29 Nov 2017 11:32:32 +0100
To: public-webrtc@w3.org
Message-ID: <5b6706b1-466b-0f47-a750-0d21dd7db9e6@alvestrand.no>
On 11/29/2017 06:27 AM, Cullen Jennings (fluffy) wrote:
> I would be fine with defining *two* experimental priorities of say 
>
> parameters.encodings[0].qosPriority =
> parameters.encodings[0].bandwidthPriority =
>
> but if don't think the experiment can redefine 
>
> parameters.encodings[0].priority = 
>
> to mean something different than in means in the non experiment case. That would become an X- prefix problem for turning off the experiment. 

I thought that at first, but found when I worked through the states that
adding an extra control gives no value.
There are 16 states (4x4) we want to reach, and with 2 controls we can
reach all of them - 4 can be reached in 2 different ways (by having
networkPriority unset, or having networkPriority set to the same value
as priority).

Having 3 controls (qosPriority, schedulePriority, priority) gives 64
states, which means that each of the 16 states we want to reach can be
reached 4 different ways - not an improvement.

My personal thought is that we'll end up with only 4 states we want to
reach (the 4 designated by setting "priority" only). But I don't think
we have data to be sure of that, so I want to enable the
experimentation, as requested.

> ( as a side note, network priority seems like a name that could cause confusion) 

only if there's something other than DSCP-rules to confuse it with. Is
there?


>
>
>
> ( Just to be 100% clear, this is send with my contributors to WebRTC hat on an no others )
>
>
>
>> On Nov 28, 2017, at 10:15 AM, Harald Alvestrand <harald@alvestrand.no> wrote:
>>
>> Picking up on a post-Singapore action item:
>>
>> I've written a very short (VERY short) spec for an extension to
>> webrtc-pc that allows one to control the setting of packet-level
>> priority separate from queue-management priority.
>>
>> This is at https://github.com/alvestrand/webrtc-dscp-exp
>>
>> Best starting point is probably the explainer:
>>
>> https://github.com/alvestrand/webrtc-dscp-exp/blob/master/explainer.md
>>
>>
>> The question now is - what now?
>>
>> Possible actions include adopting this in the WG, asking for adoption as
>> a WICG spec, or keeping it as an individual contribution.
>>
>>
>> What do people prefer?
>>
>>
>> Harald
>>
>>
>>
>

-- 
Surveillance is pervasive. Go Dark.
Received on Wednesday, 29 November 2017 10:33:06 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 29 November 2017 10:33:07 UTC