Re: [httpstreaming] [dispatch] [conex] Q-HTTP

[Mikael Abrahamsson <[e-mail]>]

On Thu, 11 Nov 2010, GARCIA ARANDA, JOSE JAVIER (JOSE JAVIER) wrote:

> Service providers are worried about ARPU. It is decreasing becasue the 
> "exaflood" phenomenon. The exponential traffic can not be sustained by 
> the network, with incremental increases in bandwidth.

I don't get it. Are you saying that because there is more traffic, the 
user is paying less money per month? Yes, profit per customer might be 
down, but why should traffic volume decrease revenue?

> These ISP capabilities can be priced to developers/content providers, increasing
> ISP revenues. Capabilities such as location, presence, billing, security, QoS....

I agree that an ISP can be a micropayment provider and also provice some 
location information.

> One of the most important is QoS. If developers can not find profitable 
> business Models, innovation is compromised. QoS means a mix of traffic 
> engineering + priorization + etc

Packet prioritization is only of value when the network is full. QoS is 
only of interest when BE works badly.

> Now imagine an ISP which offer "intelligent" QoS ( based on Q-HTTP) to enable virtualization of games
> (like www.onlive.com, but using the network instead locating servers at last mille)

I don't get this either. You can't play an FPS with tens of milliseconds 
of network delay, so you need to locate servers close to the customers to 
keep latency low, plus you also don't want the access latency to eat up 
your latency budget so ADSL and cable goes out the window anyway, the only 
thing left is the sub-millisecond latency of ETTH.

Btw, I think Q-HTTP is a horrible idea. It seems require a lot of state in 
the network. State is expensive. What happened to KISS principle?

-- 
Mikael Abrahamsson    email: [e-mail]