Re: [AVT] I-D ACTION:draft-ietf-avt-tfrc-profile-06.txt

On Sep 18, 2006, at 5:33 AM, Colin Perkins wrote:

> On 15 Sep 2006, at 17:54, Ladan Gharai wrote:
>
>> On Sep 15, 2006, at 8:18 AM, Colin Perkins wrote:
>>
>>> On 8 Sep 2006, at 22:13, Ladan Gharai wrote:
>>>
> ...
>
>>>> The draft also discusses the issues of RTCP bandwidth for AVPFCC  
>>>> flows with relatively small (say less than 20 ms) round trip  
>>>> times. The draft recommands abiding by the AVP 5% constraint,  
>>>> however this can preclude certain flows with small RTTs.
>>>>
>>>
>>> We have mechanisms for signalling a different RTCP bandwidth  
>>> fraction [RFC 3566]. I'm curious why they're not used, rather  
>>> than precluding certain flows from using congestion control?
>>>
>>
>>     my text was badly worded
>>
>>     we can certainly use mechanisms that signal using a higher  
>> RTCP bandwidth on feedback
>>
>
> Okay - the draft could do with a reference to RFC 3566.

    Certainly. I will add a reference (and discussion) to 3566 in the  
next revision of the draft.
>
>
>>     but what is the breaking point for too much feedback? 5% 10%  
>> 15% ... 50%?
>>
>>     for TFRC at small RTTs with RTCP packets things can get out of  
>> hand quickly
>>
>>     while this is more a TFRC mechanism issue (does anyone know  
>> how dccp addresses this?) I do believe that the AVPFCC profile  
>> should provide some guidelines
>>
>
> What would be the size of the feedback flow if using TCP or DCCP?  
> Being not worse than those transports would be a good start.

     I dont believe there is a problem for TCP,  since it is ack  
based. In TCP feedback is indirectly congestion controlled too, as  
halving the send rate will result in less feedback (ack) packets as  
well.

    Not sure what dccp does for CCID3 and 4?

>
> Cheers,
> Colin
>

_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
https://www1.ietf.org/mailman/listinfo/avt