Re: [httpstreaming] [AVT] Fw: Agenda and Slides

Sent from my iPad

On Nov 7, 2010, at 8:36 AM, "Colin Perkins" <csp@csperkins.org> wrote:

> On 8 Nov 2010, at 00:15, Ben Niven-Jenkins wrote:
>> On 7 Nov 2010, at 16:04, Ali C. Begen (abegen) wrote:
> ...
>>>> From: avt-bounces@ietf.org [mailto:avt-bounces@ietf.org] On Behalf Of Ben Niven-Jenkins
>>>> Colleagues,
>>>> 
>>>> Reading the slides I'd like to make some comments in advance of the bar-bof, we can discuss them more via the mailing list
>>>> or in the bar-BoF itself.
>>>> 
>>>> HTTP_Stream_1.ppt Slide 14:
>>>> 
>>>> "
>>>> No distinction regular HTTP traffic from HTTP Streaming traffic
>>>> Disadvantage:
>>>> Transport streaming media in the same way as web page
>>>> transport Streaming media has no priority to be delivered/processed first
>>>> "
>>>> 
>>>> This is not correct, it is possible to apply different treatment to HTTP Streaming traffic Vs "regular" web page traffic, e.g. by
>>>> the server setting different TOS/DSCP for streaming Vs "web" traffic.
>>> 
>>> If the network will not respect to these code points (which is the case in the open Internet), this won’t help but the servers themselves can prioritize anything they want to in their scheduling or processing. But, I am having difficulty in understanding why this is relevant to a standardization work. It looks to me as a product feature differentiation.
>> 
>> Agreed. IMO this is purely a deployment/implementation issue and not something that needs any standardisation.
> 
> It *might* make sense if the aim were to add an HTTP header to mark HTTP streaming requests, so that a web proxy treated them differently? (i.e., differentiation at the HTTP service layer, not at the network layer).
> 

But why ? It's not clear to me at all that "HTTP Streaming" traffic (whatever that means) should have 'higher priority' (whatever that means) than other web traffic. HTTP streaming clients generally have at least tens of seconds of buffer, meaning they can absorb delays in delivery, whereas for a web page every delay translates into the user waiting, and the user really wants their whole page as fast as possible.

If there are additional HTTP features that would be good for HTTP Streaming (and I am sure there are) these should be designed as generic HTTP enhancements.

This discussion needs to be much clearer about the assumptions and what is actually meant by the various terms being used.

...Mark 

> Colin
> 
> _______________________________________________
> httpstreaming mailing list
> httpstreaming@ietf.org
> https://www.ietf.org/mailman/listinfo/httpstreaming
> 