The following text is copyright 1998 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

The elusive goal of counting

By Scott Bradner

Once upon a time when the 'Net was young people thought they knew how big it was, at least in the dimension of traffic. Merit, the organization that managed the NSFnet for the National Science Foundation, used to publish monthly traffic reports. These reports listed the amount of traffic that entered and exited the NSFnet backbone at the exchange points with the regional networks.

The Internet of those days primarily consisted of a set of regional data networks, sort of geographically constrained Internet service providers (ISPs), serving customers and using the NSFnet to exchange traffic among themselves. This simple Internet architecture meant that these reports gave a reasonable idea about what was going on. Even then it was hard to use these reports to tell what the pattern of traffic exchange was since the reports only listed traffic in and out of the edges of the NSFnet and not what paths this traffic was taking through the backbone.

Those days of a simple Internet are long gone. There is no longer one backbone, there are a dozen or more depending on one's definition of a backbone. The ISPs no longer are restricted to a specific territory. There are many ISP to ISP connections and these connections for a semi-random mesh rather than a clean hierarchy. And the ISPs consider their traffic statistics to be proprietary information.

So we have no real traffic data and even if we did it would be very hard to understand the effect of the patterns of traffic. For example if I were going to send data between two sites on different ISPs in Boston that data might be able to never leave Boston if the two ISPs were interconnected locally or it might have to go through Washington DC if the ISPs only interconnected at the MAE East exchange.

That means that it is impossible to answer a question that gets asked all the time -- what are the relative traffic loads of the Internet and the telephone network? Because of FCC reporting rules there is reasonabally good data about what is going on in the phone networks but nothing but speculation about the Internet side.

There is a current reason to worry about this lack of an ability to understand just what is going on in the Internet. One of the issues in the WordCom / MCI merger proposal is the fear that the resulting company would dominate the Internet business. In the past MCI has made extravagant claims about the percentage of Internet traffic that flowed through its network. These were claims that no one could refute because there was no public data that could be used to analyze the claims. The charges of potential dominance and the defenses of limited dominance are currently only bluster because there is no public data to back them up.

It just might be time to figure out a way to get some real information about what is going on in this infrastructure that is every day becoming more vital to the economic health of the world.

disclaimer: Harvard's claims are real, not extravagant, in any case I developed the above desire for data on my own.