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

[Mikael Abrahamsson <[e-mail]>]

On Wed, 17 Nov 2010, Benjamin Niven-Jenkins wrote:

> My point was that if you look at global transit prices they are 
> typically a few dollars a Mbps in the west. That pricing is not 
> universal across the globe.

That is true, but the reason for transit prices being high is usually a 
political problem, not a technical one.

No matter of talk about how nice QoS is hides the fact that if you're 
congesting anything else than the customer port, you're not giving the 
customer what he or she purchased. Also, the Internet became successful 
because it adhered to "keep it simple stupid" principle. The suggestions 
I've seen so far here is anything but simple and usually involves quite 
intrusive changes in equipment and business models.

I totally fail to see how making the network and payment models more 
complicated will make the network better. Money (capex and opex) should be 
spent on upgrading capacity and keeping the network simple, not trying to 
make the impact of too little capacity less noticable.

The network that has constrained access bandwidth (for instance mobile 
networks) already have advanced per-user queuing that assures "fair 
access" to the media. I would imagine cable networks have the same.

I see little reason to implement support for this in end-hosts, end-hosts 
should be greedy because that's how things work in real life. End-hosts 
are under the control of the end-user and thus can't be trusted to do "the 
right thing" when it comes to "fairness".

So anyone who wants to underprovision their network need to make sure they 
have 3GPP style bearer and scheduling concept working in their network to 
handle how resources are handled, we don't need new mechanisms for this, 
it's already been available for 10 years.

-- 
Mikael Abrahamsson    email: [e-mail]