A Practical Navigator for the Internet Economy


We interview Dave Van Allen, CEO of FASTNET, a regional ISP in with operations currently in New Jersey and Pennsylvania. While FASTNET offers dial-up access, its business model focuses on providing reliable leased-line and web-hosting services to corporations. Being multi-homed to four different national service providers, it says to business customers that it can route around any failures at the national level. For mission critical applications it suggests that its customers take two FASTNET connections where each connection uses a different LEC service such as point-to-point leased line, or frame relay or SMDS. Van Allen says that he can save his customers money by multi-homing to the nationals on their behalf. He also points out that if FASTNET tried to become a national provider itself, this model would break down.

Moving on to discussions of the regulatory environment, he points out that many small ISPs don't pay adequate attention to those aspects of the business environment within which they will have to operate. He explains the rise and fall of Bell Atlantic's Centrex Extend service in Pennsylvania - noting that the disappearance of this tariff may be causing significant problems for ISPs which had counted on its availability to bring virtual pops to customers.

Van Allen is optimistic about the effect on the well-managed ISP of the competitive climate created by the 1996 Telecom Deregulation Act. He points out that he will be able to pull connections across LATA boundaries more cost effectively and adds that he is already getting some of his dial up service, NPA and NXX (area code and first three digits) from a Competitive Local Exchange Carrier (CLEC).

He warns strenuously against ISPs assuming that they can survive by doing only 20 dollar a month dial-up. He also asserts that those who discount the 20 dollar figure significantly will likely find that they do not have the capital to replace analogue dial-up technology with digital (PRIs plugged into Ascend MAX 4000's and the like). This, he believes, will be necessary in order to avoid major price increases that the RBOCs are likely to win against ISPs that continue to pull ordinary phone lines from the LEC's Central Office. We shall be doing some additional ISP business model interviews. The next will be with Frontier Net.


From the NANOG list we have edited what is the best discussion we have yet seen of the state of this critical and murky issue. Calling for a clearer demarcation between peering and transit, William Simpson began the discussion by adding "the current technology used at the exchange points could encourage abuse. What is to stop anyone connected to an exchange from simply dumping packets anonymously at the link level into the various inter exchange providers' routers and getting free transit?" In the long discussion that followed this kind of cheating was acknowledged by some of the participants to be a potential or actual problem.

Sean Doran also added a long but very well argued technical description of how BGP might be evolved to function as a peering tool. According to Doran: "Finally, all of this leads me to conclude that on a purely technical level, the denotative difference between "peer" and "transit" is so obscured as to be meaningless, except that perhaps with current tools, "transit" is easy to configure and "peering" is not. I would prefer, therefore, not to think of these as technical terms at all, but rather economic ones. This of course leads to the next thing:

Economically, the definition of "peering" vs "transit" is easy. The former is currently free, while the latter is not. Alternatively, the latter is what one sells to customers, the former is what people would like to sell, but can't figure out how to price properly. . . ..

Until recently, the bandwidth available at an exchange point was impossible to limit; an aggressive sender could swamp a particular provider's connection, as could an aggressive receiver. Moreover, as people have been discussing, without the assistance of an Exchange Point operator, there is no way to prevent people from sending traffic into your network without your permission"

One reason that the peering versus transit issue is confusing to the layman is that, when an ISP becomes a customer of an upstream provider, the monthly fee paid for the ISP's backbone connection includes a promise to deliver its packets essentially anywhere they need to go within the global Internet. However, if that same ISP goes to a network exchange point, conditions change totally. Now routing packets to another backbone to get them to their intended addressee, without a peering or transit agreement in place with that other provider, is considered theft of service from said provider.

From observing this discussion it becomes clear that getting peering with the major providers is becoming increasingly impossible to accomplish - in part, but only in part because the majors are shifting more and more of their traffic to private exchanges. Membership at the exclusive top level of the Internet seems to be increasingly restricted.


We publish a listing of international Internet links that land in the US. The listing is an attempt at a catalogue by a person knowledgeable in this area. We have been able to do some independent verification of Asian links that shows a general trend of very large increases in bandwidth there within the past year, something that we also suspect, but have not yet been able to verify, is true for Europe.


At the appropriate moment before the election, the Gore/Clinton "team" has unveiled plans to subsidize the building of a next generation research Internet for American universities. While the exact shape of what is to be built is still not clear, we examine some of the trial balloons that were floated as long ago as last April.


In the eyes of some the performance of JVNCNet, an NSFnet regional that once served R&E clients from Boston to Philadelphia, but has since shrunk primarily to customers in New Jersey, has drastically declined. JVNCNet has recently been found announcing routes belonging to a California ISP - a error that is not unprecedented, but one, in this case, where no one from JVNC would own up to the mistake. We question Princeton University's continued reliance on JVNC for its Internet connection. Princeton spokesman Ira Fuchs had no comment.



Private Interconnects as New Apex of Internet pp. 1- 10

In April of this year we reported the establishment of the first private exchange points between MCI and Sprint. Since then UUNET and BBN have joined Sprint and MCI in opening private exchange points or interconnects between each others' networks. Each major provider seems to have between two and six exchange points with every other provider. ANS has also joined the first four and will soon have private exchange points operational with all of them.

There appear to be valid network engineering reasons for opening these private interconnects. One of the participants told us that just one such interconnect enabled it to remove a continuous 20 megabit per second stream of data from an over crowded public exchange point. Nevertheless, the private interconnects have in the past six months become a new Apex of the Internet leaving the five major public North American exchange points that were formerly the top of the stratified internet in a position of diminished priority. At the now secondary level of the large public exchanges, those who had peering agreement with the new top five may very well continue to peer with them but those who did not now find it generally impossible to initiate new agreements.

New national backbone providers have emerged who can meet the 45 megabit per second connect requirements at three of the five major exchanges that would have been good for peering with Sprint or MCI in April. But only six months later they find that the rules have changed. There are now unspecified requirements for private interconnects. A problem that some of these new national providers claim to be finding in the case of the top five is that the local loops from their backbones to the public exchanges are overcrowded. They maintain that a negative consequence of this is that large commercial customers that they have won, and on which they are depending to become large enough to justify private interconnects with the top five, are complaining that access to the customers of the big five is sluggish. They find that it only takes so much sluggishness before those customers disconnect and move directly to the top five. Meanwhile the top five say that they will upgrade their connections to the public exchanges just as soon as they get their private interconnects under control. But, since backbone traffic is exploding (for example a 400% increase on BBN's backbone since January 1996), no one seems to know quite how soon this will be.

We have culled much of this information from a discussion on the NANOG list where one customer complained about packet loss and some serious debate ensued as to what part of the internet's architecture was responsible. However half of our lead article is the result of a private discussion that we facilitated between Sean Doran (Sprint alumnus) and Mike Trest (Chief Scientist of ATMnet). This discussion emphasizes network growth in Europe and describes the problems of keeping a large public exchange afloat from the point of view of technology and the impact of failure on the network itself. Finally it discusses whether the public exchanges might sprout their own backbone and create a secondary entity large enough to gain peering with the top five.

A Rural ISP Business Model, pp. 11-16, 24

We interview Johnathon Dasteel, the owner of Frontier Internet of Durango Colorado. Having started with an emphasis on dedicated 56 kbs sales in May of 1994, Dasteel realized, by the fall of that year, that meeting dial-up demand would be the only means by which he could place an adequate foundation beneath his business. He uses a fifty cent per hour charge for those who exceed fifty hours per month to finance strong customer support - something he finds very valuable in a small community where everyone knows everyone else.

This year, with US West having sold out its local frame relay availability, he has begun to experiment with wireless delivery of his local loops for dedicated customers. He has also just purchased an ISP across the state line in northwestern New Mexico. By February he will have joined the NOCs of the two ISPs by a high speed wireless link and will be dual homed into both Tech Net and the Colorado Internet Co-op.

In the face of deregulation he believes that rural communities had best be prepared to meet their own telecommunication needs. He sees the future of the rural ISP as focused on providing general telecommunications and computer systems integration services to small businesses who, lacking their own expertise, cannot afford to hire one of the large national firms.

We conclude with a Letter to the Editor from Interport describing its own experience and supporting the FASTNET business model outlined last month.

Credibility in the Second Tier of NSPs pp. 17-19

Wild West shoot outs rather than "corporatization" mark at least part of the secondary tier of National Service Providers. AGIS inks a deal with ALLTEL and the press release claims that AGIS is one of only 6 global providers "that have management control of the Internet routing tables, which are the directories that identify the location of all Internet users." AGIS and CAIS report exchange of threats against each other on NANOG mail list.

CPSR's Troubles Continue pp. 20 - 21

On October 18 Seattle based telecom policy analyst Peter Marshall published to the net a report on CPSR that provided detailed information about grants made to Doug Schuler and departed Executive Director Audrie Krause's refusal to endorse a pass through grant requested by former board member Terri Winograd. We republish much of Marshall's article and conclude with a critique of the current board's efforts to put the organization's problems behind it by means of an ad hoc committee headed by new board member Nathaniel Borenstein that is to review and report on the organization's problems.

What is a Viable Business Model for Web publishing? pp. 19, 22

We publish a discussion between ourselves, Mark Stahlman, and Chris Cooper of Quote.com. Stahlman believes that ad revenues as a major source of support for Web publishing over the long term are not viable. We point out that ad revenue assumptions are carrying forward the mass market industrial age paradigm of publishing and add that the web is much better designed to support niche market publications like our own which, with very low overhead, can prosper on a very small subscription base rather than need, like Microsoft's Slate, 50,000 subscribers to break even.