Google wants your Internet to be faster

Sean Donelan sean at donelan.com
Tue Aug 10 11:35:35 UTC 2010


On Mon, 9 Aug 2010, Christopher Morrow wrote:

> On Mon, Aug 9, 2010 at 3:18 PM, Zaid Ali <zaid at zaidali.com> wrote:
>> The devil is always in the details. The Network management piece is quite
>> glossed over and gives a different perception in the summary. You can't
>> perform the proposed network management piece without deep packet inspection
>> which violates every users privacy.
>
> how is that though? you COULD do something odd like say: "Anything to
> zaid-ali's netblocks is preferred in queues over things to
> jolymacfie's netblocks. that wouldn't require any DPI at all, just a
> traffic classification engine on/near the endpoint, say like on the
> DocSIS modem, or on the handset itself... many handsets are unix-ish
> things with some ability to do 'firewall' things, certainly they could
> mark packets outbound, certainly at peering points a network could
> classify in simple ways and mark packets properly there as well.

Or even simplier, sell seperate TDM circuits.   Although some people think 
there is only a single network, Internet; in practice the Internet has 
always been just one of many different networks built on top of various 
telecommunication facilities.

Is there a performance difference between the Internet and Internet2? 
Should that be allowed, or must all IP networks have the same performance?





More information about the NANOG mailing list