Subject: Re: The Great Exchange Date: Fri, 29 May 1998 09:56:15 -0400 From: "Perry E. Metzger" <perry@piermont.com> [...]
You appear to be confused, or perhaps confusing, on a couple of points.
The business case in all telecom markets is going towards distance insensitive pricing, and even flat rate pricing.
This isn't particularly true. The cost of a phone call shows distinct changes in pricing based on geography: from my free local calling area to intra-state to inter-state to international. You are correct that the pricing within one of these four "bands" (for lack of better term) is becoming less sensitive to distance. However, there are distinct differences in prices between "bands." Once you start using the phone analogy, you might consider pricing based on several zones: o Time-insensitive, distance-insensitive communications analogous to a toll-free, unmetered local calling area. Is this traffic that uses a local Internet exchange? o Time sensitive, distance-insensitive communications analogous to interstate phone calls. Is this usage sensitive pricing for Internet services? o Time sensitive, distance-sensitive communications analogous to international calls. Is this pricing for overseas Internet traffic for countries that have significant differences between the cost of intra-country and inter-country services? It should also be noted that the phone companies have trying to eliminate unmetered (on the basis of time) local phone service. I believe that there are fewer and fewer areas with unmetered (based on time) local phone service. So, I think a closer examination of your example may yield a different result.
The cost of providing service is very low, and not particularly traffic or distance sensitive in reality. [...]
Again, I don't know whether you are confused or confusing, but your statement ignores the differences between the average cost of providing a service and the incremental cost of providing a service. The incremental cost of forwarding a packet is obviously very small. However, the cost of operating a large, high-quality ISP is very high. We seem to expect an ISP to have, for example, a 24x7 NOC. Someone has to pay for that expense. That someone is generally the ISP's customers. The only question is how to allocate the cost of a NOC to the customers. It probably seems reasonable to charge "larger" customers more than "smaller" customers. It also seems reasonable to allocate this cost on the basis of some other variable cost, such as the size of their Internet connection, or (oh no!), the amount of traffic they generate. [I want to be there when someone tries to use this "the [incremental] cost of providing service is low" argument at the rental car counter to get a rental car for the cost of the gasoline. If someone volunteers at the next NANOG or IETF, I would like to come watch.] -tjs