A Practical Navigator for the Internet Economy
Internet Business models: cooperative, best effort or contractually defined service metrics? Settlements for my using more of your resources than you used of mine? Charges by packets or other advantages derived? Charges for traffic termination? Some of the big players have serious ideas about reshaping the economics of the internet. We begin a survey of the major player's positions. Tonight we publish two interviews. One with an a major figure who insisted on anonymity. The second with Vint Cerf on MCI's views. In October we shall publish an interview with Bill Schrader speaking for PSI. We shall also publish another interview with the senior backbone engineer for a top tier provider. Summaries of these interviews and our other articles follow.


Some of the top tier providers of the Internet have become very interested in business models. The way the top eight to ten providers shake out these issues will determine the future use and affordability of this technology. Understanding behind-the-scenes pressures should make it possible to see where the potential fracture points in the network and to gain an idea of where pressure skillfully applied has the best chance of guiding developments toward an outcome that one favors. One of the big players approached us one these issue and insisting on anonymity led us deep into the murky forests of views of the majors on such things as settlements. What follows is a summary of the issues at work according to our informant.

People are asking: What is the purpose of the Internet? What is the business model? Is it cooperation and best effort, or legally defined service metrics? Must it change? If the predominate business model becomes one of support for mission critical Fortune 1000 business activities with demands for real time audio and video and metrics of service used to guarantee this, then some think that a cooperative, best level of service model is dead.

According to this business model it is said that providers must legally spell out their obligations and responsibilities to each other. This spelling out will take the form of traffic measurements between providers and payments to the provider if the amount of traffic entering the provider's network from the customer is greater than the amount of traffic sent to the customer's network. A series of fees or settlements will be imposed top down by the big guys on the little players in the name of a more reliable Internet. Level two providers will in turn collect settlements from level three players who can either pay up sell out or demand termination fees. However, if, rather than pay the settlements demanded, the little guys demand that their larger brethren give them termination fees or disconnect them from the net, no one can be quite sure of what will happen.

The business mix of some of the big players is radically different than that of some of the others. A) MCI exists now mainly as a transit backbone. MCI is not profitable with the current non settlement based, best effort cooperative model. B) Sprint exists as a reseller whose marginal cost clients create service problems for other majors and whose slowness to reengineer (ie increase) the infrastructure of its backbone irks its competitors. MCI, BBN, ANS would like to impose some order on sprint

Given the different business mix there is some evidence that the club of six are not treating each other's backbones fairly when it comes to settlement free transit. MCI more so than the others feels abused. Increasing loads and users will necessitate the creation of more NAPs - here and in Europe and Asia - driving up the cost of remaining a top tier provider subordinate to no one. Should settlements be imposed, they will be likely to benefit the largest most wealthy networks the most, including nets like MCI that are primarily in the transit rather than the customer game and nets that because of their size and wealth can afford to play the global role most easily.

But according to the first players we interviewed (provider x and MCI) a shift in business model to guaranteed service metrics will dramatically increase the costs of providing such service and be used by the larger players who have access to more resources to raise the cost of entry into the ISP business and drive the lower cost providers out of business. Indeed higher cost corporate providers can drive the low cost mom & pop ISPs out of business only by changing the business model in such a way that the small guy can no longer afford to play.

The question could become one of can you have both a mission critical fortune 1000 net and the vibrant low cost public Internet available today? But how you could have both and not slash the value of each by restricting the amount of the world that each can reach is unknown. Can you change the current rules openly in such a way as to not allow those who don't want the rules changed to act in their own self defense? Probably not. Can you do it by stealth in such a way that you either don't loose your share or render yourself liable to anti-trust accusation? Very difficult.

Where does it ultimately shake out? Right now some are saying that it shakes out in the rules made by the top six providers as they move to a larger number of NAPs world wide. Does the six stay as the six or does it shrink to a smaller number who no longer provide full Internet connectivity to those who are unable to meet the cost of providing predefined service metrics? "Could: If you want full routing, you provide full service become the watch word." Yet the problem for the big players is who dares to move in this direction first? Only a coordinated move can succeed and without a public forum such a move would invite anti-trust litigation.

Right now the Internet is founded on the cooperative best effort model. If that model breaks, what follows? Chaos? For the model to succeed, routes must be kept under control by using CIDR effectively. Backbones must be invested in by all in true proportion to the growth of the customer base of each. More questions: Are there rewards for being uncooperative and taking advantage? Does he who abuses the MCI backbone the most and does it the soonest make out the best?

How do the RBOCs fit?. Who are they going to go to for national backbone service? How do you possibly get small providers to participate in a forum that would in effect be putting them out of business? Of course you could argue that they could stay in business by raising their rates from the $20 a month to the 80 to 100 a month that the big players deem more reasonable. Except how do you convince them that doing so wouldn't cause them to loose market share to the holdouts. Of course if the holdouts could only get to a tiny fraction of the Internet then you'd have a very different situation.


Vint Cerf talks about the need to establish an internet business model that will all providers to recoup adequate payback from their resource investment. While shying away from calling the subject of discussion settlements, he talks about the possibility of different classes of service for those who use differeing amounts of internet resources. He says that a measurement of packets sent and received is not an adequate way to bill for usage of internet resources. Having said this he calls for a discussion process aimed eventually at trying to decide who benefits from what kind of transactions and then deciding how to charge other providers for their use of your resources. While we are not comfortable with them, his positions seem reasonable. Nevertheless the direction is clear. In about a year look for major changes away from the internet model of best effort service and towards a finer granularity of charging for resources consumed according to network process used in addition to the size of the pipe attached. We also anticipate publishing some differing views on business models and settlements from one or more of the other major providers.

Vint sent us the following reply to our summary:

1. settlements usually refers to exchange payments between service providers, not to charging among users. I think the text above confuses this point. I said in our interview that the notion of transit fees in the cases that traffic did not enter or leave our network from/to a customer seemed one way to think about dealing with the "tragedy of the commons" problem.

2. If differential charges are made for different classes of service, it is still entirely possible that a charging model for "best efforts" service continues to be rather flat while showing more sensitivity for the high resource services. (Here I am thinking about various real-time services that really do consume a lot more forwarding or bit-carrying capacity).


Interviewing some SprintLink staffers who have recently left, we try to ascertain some of the reasons for their attrition. (We note also that the Sprint NAP has had no fewer than four PIs - managers- in the past year.) SprintLink, having started as a bottom up guerilla organization, is still having trouble getting adequate resources and headcount allocated to it at the highest levels of the cooperation. SprintLink has only recently become a fullfledged business product of Sprint. SprintLink is lean and mean said one of our sources. So lean and mean that it is now on an IV drip and is in danger of starving to death. On the other hand people inside of SprintLink assured up that head count was being freed up, that getting a replacement for backbone guru Vadim Antonov was a matter of highest priority and that from the network engineers point of view SprintLink was the best major Internet player to work for because it didn't micro-manage them. No matter how bumpy the current ride SprintLink is not going away any time soon. It announces 8,600 routes from MAE East - MCI holds second palce with 7,900.


Acting in secret, the managers for the state's preferred providers (Ameritech and Sprint) told winning community networks that they may get connected by these companies at no cost - if they can recruit enough buyers for their services in their respective communities. For the first time local businesses - in violation of the program's previous standards - were declared eligible for the special prices. When the news leaked, they removed us from their mail list then reopened it in moderated form in order to screen out all dissent. Their move to censor failed however because we were able to get a copy of the list members and are running our own uncensored version.


With permission we reprint a Risks post by Phil Agree on privacy risks of smart highways, including a Washington state study praising their surveillance value. We publish a letter complaining about authetication problems with domain names. Cracks are appearing in the NET99/AGIS facade. NET99 likely to bereplaced by AGIS backbone. We offer our own intro to and summary of the on-going debate over the problems engendered by the non portability of IP numbers.


Interview with Al Huff out-going director of WSIPC and citzen critique there of.