The real bandwidth numbers.

Adam L. Beberg beberg@mithral.com
Sun, 27 Jan 2002 08:33:30 -0800 (PST)


On Sun, 27 Jan 2002, Eugene Leitl wrote:

> On Sun, 27 Jan 2002, Adam L. Beberg wrote:
>
> > Enduser per-GB rates are coming.
>
> P2P with good local node discovery algorithm typically generates only
> local traffic. I.e. ISP doesn't have to spring for routing arrangement,
> and the customer gets excellent QoS (due to local bandwidth glut). So it
> should be in the best interest of customer and ISP both to distribute e.g.
> a hacked Mojo Nation which limits its scope of operation to e.g. for
> T-Online to *.dip.t-dialin.net. Special peering nodes for P2P traffic
> could be arranged, with dynamic throttling, to keep incoming and outgoing
> roughly balanced.

I can say with a great deal of personal experience that noone wants that.

The first thing people would do is modify the client and remove the locality
awareness to get around the throttling for one. Then some random port
usage, HTTP tunneling because you have to... etc etc...

> Of course, it is easier to make the customer bleed for every
> incoming/outgoing packet. However, we're approaching a time where a
> person will purchase a plug and play (no configuration, automatic local
> network discovery) *affordable* wireless router (you did get that
> message on IP about wireless cell bandwidth to be pushable to 10 GBit/s,
> and this is not even line of sight laser?), and become a bridge/node in
> urban ad hoc wireless.

Approaching sure, soon no. The ISP's are bleeding now.

- Adam L. "Duncan" Beberg
  http://www.mithral.com/~beberg/
  beberg@mithral.com