The decision to require TCP/IP was not an obvious one. First, Jennings chose a nonproprietary rather than a proprietary protocol such as DECNET. Second, among nonproprietary protocols, his choice of TCP/IP allowed the subsequent expansion of linkages to NSFNET.
The decision by Dennis Jennings, NSF’s program director for networking, to choose TCP/IP over X.25 as the mandated protocol for the emerging NSFNET was a far-reaching decision that was to have major consequences. It led almost directly to the establishment of the system of specialized, private academic networks we have today, rather than to reliance by the academic and research community on the public, commercial networks that are the mainstays of the business world. ... The choice of a proprietary protocol such as DECNET would have led to a far different networking universe than the one we have today. (Mandelbaum and Mandelbaum, 1993: 62)
These specific decisions were made within a larger context at NSF, an environment favoring widespread, open access to computer-based communication. That environment was not necessarily shared by other government agencies, as Steve Wolff pointed out: “NASA and DOE wanted to limit the network to a defined constituency, such as researchers at the federal labs and grant recipients, and run it for them. But I thought the money wouldn't be there forever, and even if it was, we should do more things with it, build bigger networks” (SRI interview with Steve Wolff, April 23, 1996). Hans-Werner Braun, co-PI of the Merit proposal, said of Wolff, “In fact, shortly after Steve Wolff started at NSF he made comments to me, and I assume to others as well, saying ‘I do not want to see this network only as a supercomputing center network’” (Merit, 1995: 17).
In response to the Connections solicitation, NSF received innovative responses from what would become two of the major regional networks: SURANET and NYSERNET. They proposed a regional, distributed network design rather than one with all universities independently connected to the regional supercomputing center (a “star” design). SURANET was a consortium of 37 southern universities. NYSERNET, based at Cornell, would be completely responsible for funding, industry relations, network equipment, and monitoring (Mandelbaum and Mandelbaum, 1993: 62).
The NYSERNET and SURANET examples caused a major paradigm shift at NSF. Instead of funding institutional connections to supercomputer centers, the NSF shifted to funding connections of ‘cohesive’ regional networks. ... NSFNET is not a network. It is an internetwork-i.e., a network of networks, which are organizationally and technically autonomous but which interoperate with one another. (Mandelbaum and Mandelbaum, 1993: 65)
Within a year after NSFNET’s beginning, the flow of traffic was so great that it was obvious an upgrade was necessary. Steve Wolff, by then the second program manager for NSFNET, headed the project. The solicitation was drafted by Wolff and Dan Van Bellegham. It called for upgrading the NSFNET backbone to 1.5 Mbps; it encouraged participation by private companies; it incorporated the three-tier structure; it required the use of TCP/IP; it provided for a phasing out of NSF support over time; and it incorporated NSF’s Acceptable Use Policy[71] (AUP), which spelled out conditions for commercial use of the network (Merit, 1995: 18-19). According to Vinton Cerf,
Among the most critical decisions that NSF made was to support the creation of "regional" or "intermediate-level" networks that would aggregate demand from the nation’s universities and feed it to the NSFNET backbone. ... The regional networks quickly became the primary means by which universities and other research institutions linked to the NSFNET backbone. NSF wisely advised these networks that their seed funding would have limited duration and they would have to become self-sustaining. (Cerf, 1995)
Robert Kahn observed that the growth of the regional networks was critical to the success of the Internet, and that this growth was partially stimulated by the phaseout of ARPANET in 1988. He noted further that 1988 marked the introduction of MCI Mail into the Internet and the commercialization of the regional networks.
The move toward the commercialization and further growth of the Internet was triggered by the Federal Networking Council (FNC)[72] decision to allow a shift toward "for profit." The NSF played the dominant role in this decision. (SRI interview with Robert Kahn, April 22, 1996).
Steve Wolff, working within a supportive environment provided by NSF Director Erich Bloch, was a key figure in these decisions. Wolff saw that commercial interests eventually had to become a part of the network and provide financial support for it:
It had to come, because it was obvious that if it didn’t come in a coordinated way, it would come in a haphazard way, and the academic community would remain aloof, on the margin. That’s the wrong model: multiple networks again, rather than a single Internet. There had to be commercial activity to help support networking, to help build volume on the network. That would get the cost down for everybody, including the academic community-which is what the NSF was supposed to be doing. (Merit, 1995: 33)
Vinton Cerf sees Wolff as the “absolute key” to the policy decision to allow for-profit information providers on the network (SRI interview with Cerf).
Wolff and others pointed to Bloch’s role: “The unsung hero of the project is Erich Bloch; his support never wavered” (SRI interview with Wolff). “Erich Bloch joined the NSF and fostered an environment for networking” (SRI interview with Kahn). Erich Bloch knew how to get the money and play the politics; Steve Wolff knew what to do with it” (SRI interview with Kahn).
NSF's decision to permit commercial access to NSFNET (under conditions specified by the Acceptable Use Policy) and to provide support for the backbone under conditions that were intended to lead to privatization of the Internet generated controversy both within and outside the Foundation. During congressional hearings on NSFNET in 1992, a number of concerns were identified:
Rick Boucher, Chair of the Subcommittee on Science of the House Science, Space, and Technology Committee requested the NSF Inspector General to investigate these and other issues that arose after the initial 1987 solicitation for management and operation of the backbone. The Inspector General's 1993 report, "Review of NSFNET" (NSF, 1993),concluded that NSF's actions were legal and within the Foundation's discretionary authority. Criticism focused on the lack of documentation of NSF's decision to allow commercial access and on the lack of review of the decision by peers, supervisory staff, or the National Science Board.[73]
The 1988 solicitation, won by Merit, Inc., paved the way for the
next major upgrade, which came in 1993, when NSF offered a solicitation
calling for a new, high-speed research backbone (vBNS, or very
high speed backbone network service), Network Access Points (NAPs),
a Routing Arbiter, and Regional Network Provider awards. The
backbone would operate at 155 Mbps; NAPs would act as interconnection
points for commercial Internet service providers; the Regional
Arbiter would manage the growing routing tables and databases
for the providers connecting at the NAPs. Regional and midlevel
networks would continue to receive NSF funding, phased out over
a 4-year period. The high-speed research backbone, vBNS, was
given to MCI; NAP manager awards were given to Sprint, MFS Datanet,
and Bellcore. Thus the basis for the current Internet was laid,
and was realized in April 1995.[74]
In this concluding section, we focus on four aspects of the evolution
of the Internet: (1) government, industry, and university roles
and relationships; (2) the relationships between fundamental research
and technology development; (3) the role of property rights; and
(4) the role of NSF. In the case of NSF's role, we distinguish
between innovative leadership and financial support.
Support can take several forms, including support for research,
for education, and for infrastructure (e.g., travel, coordination,
instrumentation, and facilities).
Vinton Cerf's brief characterization of the shift in support and control of the Internet captures the dramatic change that occurred over the innovation's 20-year history: "A fully commercial system of backbones has been erected, where a government sponsored system once existed. Indeed, the key networks that made the Internet possible ... are now gone-but the Internet thrives!" The first "Internet" was the interconnection of dissimilar networks consisting of ARPANET, packet radio, and packet satellite under Kahn's sponsorship in 1977. This and the subsequent connection of ARPANET and CSNET were entirely government owned and supported, with university researchers and contractors such as BBN supplying hardware, software, and services. Within a decade, as a result of key decisions made primarily by NSF, a three-tiered system of internetworks existed, managed and supported by a mix of universities, nonprofit organizations, and government agencies, with portions operated by common carriers and commercial service providers. At present, the mix of operators remains, but the basic Internet backbone is owned and operated by commercial firms.
A significant aspect of this institutional evolution can be attributed to NSF's insistence that, whatever form the research-based network it originally envisioned might take, it should be accessible to supercomputer users and PC users alike; it should permit seamless interconnections among all major existing networks; and, to ensure its survival, it should eventually become self-supporting. These features meant choosing TCP/IP over other protocols, working out ways for commercial activities to be included on the Internet, and building bridges between NSF and other federal agencies so that all major government networks were included. None of these decisions were easy or obvious, and the ones involving commercial use of the Internet proved especially challenging.
In the mid- to late 1960s, before ARPANET's implementation, the issue of packet switching as a technology for data networks was debated during the Spring and Fall Joint Computer Conferences. The communications industry was skeptical. SRI's Don Nielson observed that industry initially "dropped the ball" on the Internet. None of the common carriers foresaw a market for digital communication; they regarded themselves as suppliers of telephone services only. Indeed, several of our respondents commented on AT&T's extreme skepticism-even opposition-when the idea of packet-switched, rather than circuit-switched, networks first appeared. Computer manufacturers such as IBM and Digital saw the potential for profit but insisted on using their own (proprietary) protocols. Since there was no industry leader, a dominant industry protocol did not emerge. In this sense, government-first DOD/ARPA and then NSF-played the role of key innovator in the evolution of the Internet. Agency leaders and program managers took technical and political risks, and the risk-taking seems to have paid off.
All three institutions-government, industry, and universities-played significant but different roles in the development of the Internet. To some extent, this extensive involvement of all three types of institutions can be attributed to an "invisible college" that transcended institutional sectors, a feature of many areas of U.S. science and engineering. The focus of technical and organizational innovation shifted from government to industry over a 30-year period, with universities playing a constant supportive role over the entire time. One way to capture both the nature of these differing roles and their relationships is to trace the career paths of some of the major contributors to the Internet: Cerf, Kahn, Jennings, Wolff, Baran, Roberts, Mills, Andreessen, Bloch. Cerf began as an academic in computer science at Stanford, then joined ARPA as a program manager, was a Vice President of CNRI with Robert Kahn during 1986-94, and is now a Senior Vice President of MCI. Kahn also began as an academic, in electrical engineering at MIT, then took leave to go to BBN until he joined ARPA. Jennings came to NSF from University College at Dublin, and Wolff from the Army Research Office after serving on the faculty at Johns Hopkins. Baran, an electrical engineer by training, moved from industry to RAND, a Federally Funded Research and Development Center, and explored doctoral work at UCLA. Roberts worked at Lincoln Lab before joining ARPA. Mills moved from university (Maryland) to industry (Communication Satellite Corporation) and back to university (Delaware) at the time of his fuzzball router contributions. Andreessen went directly from undergraduate work at Illinois to help form Netscape. Bloch became Director of NSF after a long career at IBM.
Interinstitutional linkages are reflected in a different way in
membership on the myriad advisory committees to the Internet over
its history and in the coalitions that formed to provide network
services. Examples include Merit, Inc., a consortium of IBM,
MCI, and the University of Michigan, and the Internet Architecture
[formerly Activities] Board and its several task forces, such
as the Internet Engineering Task Force, on which government officials,
academics, and industry representatives serve. In many respects,
the Internet's existence exemplifies the extraordinary (relative
to most other industrialized nations) frequency and ease with
which scientists and engineers in the United States move back
and forth among government, industry, and university. It also
illustrates how a government agency's leadership can build on
this interinstitutional permeability to generate productive collaboration
at the national level.
The Internet appears, overall, to be primarily a problem-driven, technology-based innovation that required little direct input from fundamental research for its realization. The driving forces, interestingly, were not profit incentives in the private market, but public goods, first in the realm of national defense and subsequently in the university and government research infrastructure, as a means of fostering communication among computer scientists. What we are calling the Internet's intrinsic technologies-network design, packet switching, routers, protocols, browsers-were the products of problem-driven research conducted in universities and government contractor laboratories with government support. One possible exception is the research conducted at the University of Illinois's NCSA, which took place in an environment (according to Andreessen) that enabled researchers to head off in directions that looked "interesting" without seeking justification. Nonetheless, the context was one of application, as suggested by the Center's name.
Although the evolution of the Internet did not encounter technical
roadblocks that required fundamental research for their resolution
before further progress could be made, there is obvious, fundamental
research content in both the Internet's intrinsic and supporting
technologies. The electronic and physical infrastructures that
comprise the Internet clearly depend on information theory, solid
state physics, electro-optics, and other fields on which modern
communications technology is based and for which NSF has provided
substantial research support.
An obvious feature of the Internet is its "ownerlessness."
It is perhaps astonishing that an innovation that has proved
to be of such tremendous economic significance and that has spawned
numerous personal fortunes and substantial commercial profits
is based almost entirely on nonproprietary technology. Yet, as
we have seen, the key technical innovations were developed with
government support and, because of the vision of Vinton Cerf,
Robert Kahn, Steve Wolff, and other pioneers, decisions to use
nonproprietary technology were made whenever there was a choice.
Another unusual feature of the Internet is that most browser
and utility software, including Mosaic, Netscape, Fetch, Gopher,
and others, was usually made available free on the World Wide
Web. Mosaic was distributed free because of NSF's support of
NCSA; in the case of Netscape and most other browsers, profits
are realized not through purchase of the software by individual
users (to whom much of it is largely free) but through its selection
by commercial entities as the basis for construction and use of
their own Web sites. In contrast to many other types of innovations,
diffusion of the Internet's intrinsic technologies was enhanced
rather than inhibited by their public character. It was apparently
an advantage, from a technology diffusion point of view, that
the profitmaking aspects of the Internet were not realized until
most of its present features were already in place. By about
1990, the public-goods nature of the Internet began to be intermingled,
successfully, with commercial interests. It is beyond the scope
of this case to explore the extent to which this unusual evolution
is repeatable for other innovations. Doing so may also be premature,
in that the innovation is still evolving.
As we noted above, NSF's role (and the roles of other agencies) can be separated into leadership and support categories. With respect to leadership, it is well documented that NSF program managers Dennis Jennings, Steve Wolff, and Jane Caviness, working within the highly supportive environment provided-at least at the level of top management-by Erich Bloch, took risks[75], developed highly creative solutions to difficult problems (such as how to incorporate commercial activity into NSFNET), and provided essential services in coordinating among other federal agencies, academic researchers, and industry. NSFNET's initiation and architecture are attributable to Jennings and Wolff, and the transition to commercial support to Wolff and Caviness. As Mandelbaum and Mandelbaum observed, a different set of decisions by NSF "would have led to a far different networking universe than the one we have today." NSF was a leader among equals in various coordinating committees, such as the Federal Networking Council, in which, according to Robert Kahn, "NSF played the dominant role."
As the Charles Babbage Institute's report documents, the foundations of the Internet were laid by ARPA in the Department of Defense. But by the mid-1980s, primary financial support of the Internet had been assumed by NSF. NSF's support of education and infrastructure, although more subtle and difficult to identify, permeates the history of many of the key contributors. Some argue that the birth of the Internet occurred when two federal networks, ARPANET and CSNET, were interconnected in 1983. NSF provided the planning grant to Larry Landweber for the connection in 1980. Vinton Cerf and Robert Kahn point to substantial NSF support for their own personal education and doctoral support: NSF supported Kahn's doctoral studies in electrical engineering under a fellowship; his dissertation adviser at Princeton received partial research support from NSF. Leonard Kleinrock and the "UCLA Mafia" were supported almost entirely by ARPA; Gerald Estrin (Cerf's Ph.D. adviser) obtained some support from NSF as well. During the 1970s, Estrin advised Vinton Cerf, Steve Crocker, and, earlier, Paul Baran. NSF provided about a third of Dave Mills' support beginning in about 1982. The award to NCSA provided direct support to Marc Andreessen, as well as the environment and resources that enabled him to implement Mosaic as a browser for all computer users, not just supercomputer researchers.
NSF also provided a substantial amount of the university-based
computing infrastructure that NSFNET joined together. Aspray
and Williams (1994) show that, between 1959 and 1971, NSF made
more than 400 awards for computing facilities to colleges and
universities under the computer center facilities program, totaling
more than $60 million.[76] By the time the program was terminated,
most U.S. academic institutions had established computing centers.
ARPA support for computer science research during this period
dwarfed NSF's, but it was concentrated at a small number of selected
computer science departments. NSF support of computing facilities,
networks, and research in computer science and engineering resumed
again in the 1980s, resulting in five supercomputer centers, CSNET
and NSFNET, partial support for regional networks, and substantial
support for research and other infrastructure elements. By the
late 1980s, NSF support for computer and information science and
engineering amounted to well over $100 million annually.[77]
Aboba, Barnard. The Online User’s Encyclopedia. New York: Addison-Wesley, 1993.
Andreessen, Marc. Interview with LAN Times, Thom Stark Home Page, http://www.dnai.com/~thomst/.
Aspray, William, and Williams, Bernard O. "Arming American Scientists: NSF and the Provision of Scientific Computing Facilities for Universities, 1950-1973," IEEE Annals of the History of Computing, 16, 4 (1994): 60-74.
Cerf, Vinton G. http://www.cs.washington.edu/homes/lazowska/cra/networks.html.
Cerf, Vinton. “How the Internet Came to Be,” as told to Bernard Aboba, 1993.
Cerf, Vinton. “Computer Networking: Global Infrastracture for the 21st Century,” 1995.
Charles Babbage Institute (CBI). “An Interview with Robert Kahn,” OH 158, March 22, 1989.
Charles Babbage Institute (CBI). “An Interview with Paul Baran, March 5, 1990a.
Charles Babbage Institute (CBI). “An Interview with Vinton Cerf, OH 191, April 24, 1990b.
Comer, Douglas. “The Computer Science Research Network CSNET: A History and Status Report,” Communications of the ACM, 26, 10 (1983): 747-753.
Data Communications (March 1988): 270.
Jennings, Dennis, Landweber, Lawrence, Fuchs, Ira, and Farber, David. “Computer Networking for Scientists,” Science 23 (February 1986).
Kahin, Brian, ed. Building Information Infrastructure. New York: McGraw-Hill, 1993.
Kahin, Brian. "Overview: Understanding the NREN," in Kahin (1993).
Kahn, Robert. "The Role of Government in the Evolution of the Internet," in National Academy of Engineering, Revolution in the U.S. Information Infrastructure. Washington DC: National Academy Press, 1995.
Kleinrock, Leonard. “Issues in the Design of the NREN,” in Kahin (1993).
Krol, E., and Hoffman, E. Network Working Group Request for Comments: RFC 1462, FYI on "What Is the Internet?" (May 1993).
Mandelbaum, Richard, and Mandelbaum, Paulette A. "The Strategic Future of the Mid-Level Networks,” in Kahin (1993).
Merit Network, Inc. NSFNET, Final Report 1987-1995.
National Academy of Engineering. Revolution in the U.S. Information Infrastructure. Washington, DC: National Academy Press, 1995.
National Research Council. Realizing the Information Future: The Internet and Beyond. Washington, DC: National Academy Press, 1994.
National Science Foundation, FY 1988 Budget to the Congress.
National Science Foundation, Office of the Inspector General. "Review of NSFNET" (March 23, 1993).
Norberg, Arthur, and O'Neill, Judy. A History of the Information Processing Techniques Office of the Defense Advanced Research Projects Agency. Minneapolis, MN: The Charles Babbage Institute, 1992.
U.S. Congress, Office of Technology Assessment. High Performance
Computing and Networking for Science. Background Paper. Washington,
DC: U.S. Government Printing Office, September 1989.
Paul Baran, RAND Corporation (retired)
Leonard Kleinrock, UCLA
Robert Kahn, Corporation for National Research Initiatives
Vinton Cerf, MCI
Don Nielson, SRI International
Elizabeth Feinler, NASA
Steve Wolff, Cisco Systems
Peter Ford, MCI
Dave Mills, University of Delaware
Last update: 02/14/97