On Mon, 12 May 1997, Hong Chen wrote:
It is indeed an interesting question. If you look at from a different angle, and from UUnet point of view, you will see the why. The cost of providing a full bandwith T1 to UUNET's customer is actually a function of 'How Long the Packet Travels * bps', where bps is the speed 1.54Mbps in the case of T1.
If UUNet is carring the backbone burden, the first parameter is larger and bps is still 1.54Mbps. Thus, the cost of UUnet providing T1 will be much bigger than the cost of a Small ISP providing T1 for its customers.
Now the smaller ISP can price its T1 at lower cost, and compete with UUNet which cannot lower the price below its cost. By having the smaller ISP to pay for the peering just RAISE the cost of providing T1 by the smaller ISP and REDUCE the cost of providing T1 for UUnet.
This is true as long as UUNet is providing the service to the small ISP. If the small ISP chooses to connect to another NSP, rather than pay for peering, UUNet's costs and revenues are still the same, but the traffic it has to support for the small ISP is still the same. In fact, if the small ISP is "close" to the exchange point, UUNet has just helped one of it's competitors become (slightly) more profitable, by giving it a high-bandwidth customer at the cheapest point to connect a high-bandwidth customer. I guess I just don't see how forcing small ISPs/farms to pay for peering benefits UUNet unless the ISP/farm buys service from UUNet. Given UUNet's pricing, I just can't see that these high-bandwidth users will be purchasing from UUNet. They'll most likely go where the bandwidth/dollar is better, and only make things worse/costlier for UUNet (and other NSPs). Pete Kruckenberg pete@inquo.net