This story appeared on Network World at


NSFNET: The vibrant ghost of Christmases past


By Scott Bradner


Network World , 12/11/2007


At the start of the Christmas shopping season 20 years ago the National Science Foundation announced that a group consisting of Michigan's Merit Network, IBM and MCI had won a contract to develop and deploy the T-1 NSFNET. This network led directly to the Internet of today -- the NSFNET was a  gift that has kept on giving.


The 1.544Mbps NSFNET was not all that fast, even in those days, but was a lot faster than its 56Kbps predecessor. The traffic load on this network grew at a rapid rate -- as much as a 20% increase a month -- and the load soon outstripped capacity. A few years later the NSFNET backbone speed was increased to T-3 (45Mbps), which did help for a while, but only for a while.


We would never have had the Internet we know today if the NSFNET stayed the only network game in town. But from the very beginning the NSFNET prohibited the use of the network by commercial traffic. There was a great deal of criticism of this decision by some observers who felt it was hindering the expansion of the use of the net, but the decision was clearly the right one since it forced the development of commercial ISPs. These ISPs quickly outstripped, then dwarfed, the NSFNET in terms of capacity and connections and were easily able to take on the load when the NSFNET was shut down less than nine years after the T-1 network went into service. (go here for a brief history of the NSFNET).


But even though it had a short life, the NSFNET was a key, if not the key reason we have the Internet of today. The NSFNET showed you could build and operate a high-speed network backbone to interconnect regional networks and end sites. It proved that end-to-end communication over such a network of networks would work even at large scale. At the start of the NSFNET era there were about 10,000 hosts on the 'Net -- this had grown to over 6 million by the end of the era. Not all of these hosts interconnected over the NSFNET and that was part of what made the system so strong. There were thousands of ISPs of all sizes interconnecting over a half-dozen or more nationwide backbones by the mid 1990s -- the NSFNET had become just a part of a much greater whole.


The NSFNET also proved that TCP/IP could be used in large networks without any sort of central manager. When the NSFNET started there were many other, mostly proprietary, protocols to choose from if you were building an enterprise network. But management insisted that TCP/IP was the only protocol permitted on the NSFNET. This helped force the understanding that proprietary protocols did not enable interorganizational communication and quickly led to widespread adoption of TCP/IP.


The NSFNET itself is no longer with us, but it is good to celebrate its short life and the organizations and people that made it work -- and the Internet that it enabled. An Internet that seems to have no bounds (except the bounds that some telecom companies and governments would like to impose) and which will be used to buy 10s of billions of dollars worth of Christmas gifts this year.


Disclaimer: Harvard has lasted a bit longer than the NSFNET did, and I suspect, if asked, Harvard would say that ((it has made at least as big an impact on the world. But)) the legacy of the NSFNET is easier to spot right now. In any case, the above opinion is mine, not the university's.


All contents copyright 1995-2008 Network World, Inc.