Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
The Role of Government in the Evolution of the Internet ROBERT E. KAHN This paper discusses the role of government in the continuing evolution of the Internet. From its origins as a U.S. government research project, the Internet has grown to become a major compo- nent of network infrastructure, linking millions of machines and tens of millions of users around the world. Although many nations are now involved with the Internet in one way or another, this paper focuses on the primary role the U.S. government has played in the Internetâs evolution and discusses the role that governments around the world may have to play as it continues to develop. Very little of the current Internet is owned, operated, or even controlled by governmental bodies. The Internet indirectly receives government support through federally funded academic facilities that provide some network-related services. Increasingly, however, the provision of Internet communication services, regardless of use, is being handled by commercial firms on a profit-making basis. This situation raises the question of the proper long-term role for government in the continued evolution of the Internet. Is the Internet now in a form where government involvement should cease entirely, leaving private-sector interests to determine its future? Or, does A version of this paper appeared in the August 1994 issue of Communications of the ACM. 13
14 ROBERT E. KAHN government still have an important role to play? This paper con- cludes that government can still make a series of important contribu- tions. Indeed, there are a few areas in which government involve- ment will be vital to the long-term well-being of the Internet. ORIGINS OF THE INTERNET The Internet originated in the early 1970s as part of an Advanced Research Projects Agency (ARPA) research project on âinternet- working.â At that time, ARPA demonstrated the viability of packet switching for computer-to-computer communication in its flagship network, the ARPANET, which linked several dozen sites and per- haps twice that number of computers into a national network for computer science research. Extensions of the packet-switching con- cept to satellite networks and to ground-based mobile radio networks were also under development by ARPA, and segments of industry (notably not the traditional telecommunications sector) were show- ing great interest in providing commercial packet network services. It seemed likely that at least three or four distinct computer networks would exist by the mid-1970s and that the ability to communicate among these networks would be highly desirable if not essential. In a well-known joint effort that took place around 1973, Robert Kahn, then at ARPA, and Vinton Cerf, then at Stanford, collaborated on the design of an internetwork architecture that would allow packet networks of different kinds to interconnect and machines to commu- nicate across the set of interconnected networks. The internetwork architecture was based on a protocol that came to be known as TCP/ IP. The period from 1974 to 1978 saw four successively refined versions of the protocol implemented and tested by ARPA research contractors in academia and industry, with version number four even- tually becoming standardized. The TCP/IP protocol was used ini- tially to connect the ARPANET, based on 50 kilobits per second (kbps) terrestrial lines; the Packet Radio Net (PRNET), based on dual rate 400/100 kbps spread spectrum radios; and the Packet Satellite Net (SATNET), based on a 64 kbps shared channel on Intelsat IV. The initial satellite Earth stations were in the United States and the United Kingdom, but subsequently additional Earth stations were activated in Norway, Germany, and Italy. Several experimental PRNETs were connected, including one in the San Francisco Bay
EVOLUTION OF THE INTERNET 15 area. At the time, no personal computers, workstations, or local area networks were available commercially, and the machines involved were mainly large-scale scientific time-sharing systems. Remote access to time-sharing systems was made available by terminal ac- cess servers. The technical tasks involved in constructing this initial ARPA Internet revolved mainly around the configuration of âgateways,â now known as routers, to connect different networks, as well as the development of TCP/IP software in the computers. These were both engineering-intensive tasks that took considerable expertise to ac- complish. By the mid-1980s, industry began offering commercial gateways and routers and started to make available TCP/IP software for some workstations, minicomputers, and mainframes. Before this, these capabilities were unavailable; they had to be handcrafted by the engineers at each site. In 1979, ARPA established a small Internet Configuration Con- trol Board (ICCB), most of whose members belonged to the research community, to help with this process and to work with ARPA in evolving the Internet design. The establishment of the ICCB was important because it brought a wider segment of the research com- munity into the Internet decision-making process, which until then had been the almost-exclusive bailiwick of ARPA. Initially, the ICCB was chaired by a representative of ARPA and met several times a year. As interest in the ARPA Internet grew, so did interest in the work of the ICCB. During this early period, the U.S. government, mainly ARPA, funded research and development work on networks and supported the various networks in the ARPA Internet by leasing and buying components and contracting out the systemâs day-to-day operational management. The government also maintained responsibility for overall policy. In the mid- to late 1970s, experimental local area networks and experimental workstations, which had been developed in the research community, were connected to the Internet according to the level of engineering expertise at each site. In the early 1980s, Internet-compatible commercial workstations and local area networks became available, significantly easing the task of getting connected to the Internet. The U.S. government also awarded contracts for the support of various aspects of Internet infrastructure, including the maintenance
16 ROBERT E. KAHN of lists of hosts and their addresses on the network. Other govern- ment-funded groups monitored and maintained the key gateways be- tween the Internet networks in addition to supporting the networks themselves. In 1980, the U.S. Department of Defense (DOD) adopted the TCP/IP protocol as a standard and began to use it. By the early 1980s, it was clear that the internetwork architecture that ARPA had created was a viable technology for wider use in defense. EMERGENCE OF THE OPERATIONAL INTERNET The DOD had become convinced that if its use of network- ing were to grow, it needed to split the ARPA Internet (called ARPANET) in two. One of the resulting networks, to be known as MILNET, would be used for military purposes and mainly link mili- tary sites in the United States. The remaining portion of the network would continue to bear the name ARPANET and still be used for research purposes. Since both would use the TCP/IP protocol, com- puters on the MILNET would still be able to talk to computers on the new ARPANET, but the MILNET network nodes would be located at protected sites. If problems developed on the ARPANET, the MILNET could be disconnected quickly from it by unplugging the small number of gateways that connected them. In fact, these gate- ways were designed to limit the interactions between the two net- works to the exchange of electronic mail, a further safety feature. By the early 1980s, the ARPA Internet was known simply as the Internet, and the number of connections to it continued to grow. Recognizing the importance of networking to the larger computer science community, the National Science Foundation (NSF) began supporting CSNET, which connected a select group of computer science researchers to the emerging Internet. This allowed new re- search sites to be placed on the ARPANET at NSFâs expense, and it allowed other new research sites to be connected via a commercial network, TELENET, which would be gatewayed to the ARPANET. CSNET also provided the capacity to support dial-up e-mail connec- tions. In addition, access to the ARPANET was informally extended to researchers at numerous sites, thus helping to further spread the networking technology within the scientific community. Also during this period, other federal agencies with computer-oriented research programs, notably the Department of Energy (DOE) and the National
EVOLUTION OF THE INTERNET 17 Aeronautics and Space Administration (NASA), created their own âcommunity networks.â The TCP/IP protocol adopted by DOD a few years earlier was only one of many such standards. Although it was the only one that dealt explicitly with internetworking of packet networks, its use was not yet mandated on the ARPANET. However, on January 1, 1983, TCP/IP became the standard for the ARPANET, replacing the older host protocol known as NCP. This step was in preparation for the ARPANET-MILNET split, which was to occur about a year later. Mandating the use of TCP/IP on the ARPANET encouraged the addi- tion of local area networks and also accelerated the growth in num- bers of users and networks. At the same time, it led to a rethinking of the process that ARPA was using to manage the evolution of the network. In 1983, ARPA replaced the ICCB with the Internet Activities Board (IAB). The IAB was constituted similarly to the old ICCB, but the many issues of network evolution were delegated to 10 task forces chartered by and reporting to the IAB. The IAB was charged with assisting ARPA to meet its Internet-related R&D objectives; the chair of the IAB was selected from the research community supported by ARPA. ARPA also began to delegate to the IAB the responsibility for conducting the standards-setting process. Following the CSNET effort, NSF and ARPA worked together to expand the number of users on the ARPANET, but they were constrained by the limitations that DOD placed on the use of the network. By the mid-1980s, however, network connectivity had be- come sufficiently central to the workings of the computer science community that NSF became interested in broadening the use of networking to other scientific disciplines. The NSF supercomputer centers program represented a major stimulus to broader use of net- works by providing limited access to the centers via the ARPANET. At about the same time, ARPA decided to phase out its network research program, only to reconsider this decision about a year later when the seeds for the subsequent high-performance computer initia- tive were planted by the Reagan administration and then-Sen. Albert Gore (D-Tenn.). In this period, NSF formulated a strategy to assume responsibility for the areas of leadership that ARPA had formerly held and planned to field an advanced network called NSFNET. NSFNET was to join the NSF supercomputer centers with very high
18 ROBERT E. KAHN speed links, then 1.5 megabits per second (mbps), and to provide members of the U.S. academic community access to the NSF super- computer centers and to one another.1 Under a cooperative agreement between NSF and Merit, Inc., the NSFNET backbone was put into operation in 1988 and, because of its higher speed, soon replaced the ARPANET as the backbone of choice. In 1990, ARPA decommissioned the last node of the ARPANET. It was replaced by the NSFNET backbone and a series of regional networks most of which were funded by or at least started with funds from the U.S. government and were expected to become self-sup- porting soon thereafter. The NSF effort greatly expanded the in- volvement of many other groups in providing as well as using net- work services. This expansion followed as a direct result of the planning for the High Performance Computing Initiative (HPCI), which was being formed at the highest levels of government. DOD still retained the responsibility for control of the Internet name and address space, although it continued to contract out the operational aspects of the system. The DOE and NASA both rely heavily on networking capability to support their missions. In the early 1980s, they built High Energy Physics Net (HEPNET) and Space Physics Analysis Net (SPAN), both based on Digital Equipment Corporationâs DECNET protocols. Later, DOE and NASA developed the Energy Sciences Net (ESNET) and the NASA Science Internet (NSI), respectively; these networks supported both TCP/IP and DECNET services. These initiatives were early influences on the development of the multiprotocol net- working technology that was subsequently adopted in the Internet. International networking activity was also expanding in the early and mid-1980s. Starting with a number of networks based on the X.25 standard as well as international links to ARPANET, DECNET, and SPAN, the networks began to incorporate open internetworking protocols. Initially, Open Systems Interconnection (OSI) protocols were used most frequently. Later, the same forces that drove the United States to use TCP/IPâavailability in commercial worksta- tions and local area networksâcaused the use of TCP/IP to grow internationally. 1 For a brief period in the mid-1980s, there was a small initial NSFNET that linked the supercomputer centers with 64 kbps lines.
EVOLUTION OF THE INTERNET 19 The number of task forces under the IAB continued to grow, and in 1989, the IAB consolidated them into two groups: the Internet Engineering Task Force (IETF) and the Internet Research Task Force (IRTF). The IETF, which had been formed as one of the original 10 IAB Task Forces, was given responsibility for near-term Internet developments and for generating options for the IAB to consider as Internet standards. The IRTF remained much smaller than the IETF and focused more on longer-range research issues. The IAB struc- ture, with its task-force mechanism, opened up the possibility of getting broader involvement from the private sector without the need for government to pay directly for their participation. The federal role continued to be limited to oversight control of the Internet name and address space, the support of IETF meetings, and sponsorship of many of the research participants. By the end of the 1980s, IETF began charging a nominal attendance fee to cover the costs of its meetings. The opening of the Internet to commercial usage was a signifi- cant development in the late 1980s. As a first step, commercial e- mail providers were allowed to use the NSFNET backbone to com- municate with authorized users of the NSFNET and other federal research networks. Regional networks, initially established to serve the academic community, had in their efforts to become self-suffi- cient taken on nonacademic customers as an additional revenue source. NSFâs Acceptable Use Policy, which restricted backbone usage to traffic within and for the support of the academic commu- nity, together with the growing number of nonacademic Internet us- ers, led to the formation of two privately funded and competing Internet carriers, both spin-offs of U.S. government programs. They were UUNET Technologies, a product of a DOD-funded seismic research facility, and Performance Systems International (PSI), which was formed by a subset of the officers and directors of NYSERNET, the NSF-sponsored regional network in New York and the lower New England states. Beginning in 1990, Internet use was growing by more than 10 percent a month. This expansion was fueled significantly by the enormous growth on the NSFNET and included a major commercial and international component. NSF helped to stimulate this growth by funding both incremental and fundamental improvements in Internet routing technology as well as by encouraging the widespread distri-
20 ROBERT E. KAHN bution of network software from its supercomputer centers. Inter- connections between commercial and other networks are arranged in a variety of ways, including through the use of the Commercial Internet Exchange (CIX), which was established, in part, to facilitate packet exchanges among commercial service providers. Recently, the NSF decided that additional funding for the NSFNET backbone no longer was required. The agency embarked on a plan to make the NSF regional networks self supporting over a period of several years. To assure the scientific research community of continued network access, NSF made competitively chosen awards to several parties to provide network access points (NAPs) in four cities. NSF also selected MCI to provide a very high speed backbone service, initially at 155 mbps, linking the NAPs and several other sites, and a routing arbiter to oversee certain aspects of traffic alloca- tion in this new architecture. The Internet Society was formed in 1992 by the private sector to help promote the evolution of the Internet, including maintenance of the Internet standards process. In 1992, the IAB was reconstituted as the Internet Architecture Board, which became part of the Internet Society. It delegated its decision-making responsibility on Internet standards to the leadership of the IETF, known as the Internet Engi- neering Steering Group (IESG). While not a part of the Internet Society, the IETF produces technical specifications as possible can- didates for future protocols. The Internet Society now maintains the Internet Standards Process, and the work of the IETF is carried out under its auspices. ISSUES FOR CONSIDERATION As the Internet continues to grow, the role of the research com- munity in developing and evolving standards needs to be addressed. When the financial implications of decisions about Internet standards were relatively small, the current standards process proved entirely satisfactory. As the financial impact of such decisions becomes in- creasingly significant, the nature of the standards-setting process will continue to change to allow more direct industrial involvement. How this will ultimately play out is unclear. However, the vitality of the current process derives from the broad involvement of the many communities that have a stake in the Internet. Unlike typical top-
EVOLUTION OF THE INTERNET 21 down standards-setting operations that implement decisions formed by consensus, the Internet process works essentially in reverse through a kind of grass-roots mechanism. Candidates for Internet standards ordinarily result from actual implementation and wide- spread experimentation within the IETF. The most promising of these candidates are selected for placement on the Internet standards track. No better process has yet emerged that is as dynamic and allows as much direct involvement by industry. Further, with the widespread internationalization of the Internet, scores of countries now have fundamental interests in its evolution. Within the United States, the Internet is seen in many quarters as the starting point for the National Information Infrastructure (NII). Around the world, there is growing recognition that the set of NIIs (assuming each country commits to developing one) should be com- patible with each other along some still-unknown dimensions. Who should take the lead in ensuring this compatibility? Is this a role for the private sector, for governments acting together, or for some com- bination of the two? There is clearly a role for government, at least to provide oversight, support, and guidance, if not to participate actively. Apart from these issues is concern about the viability of any approach that has no individual or organization with overall responsi- bility for its evolution. It seems fair to say that many of the tradi- tional Internet carriers would prefer that new capabilities be provided by them as a turnkey service. Industry surely has the capacity to provide many of the necessary capabilities, but history has shown the importance of government involvement. What guarantees that the same degree of vitality will be part of its future evolution if market forces alone determine what new capabilities are added to the Inter- net? Furthermore, the Internet offers the possibility of bypassing conventional service offerings by regulated carriers. This may both make it extremely difficult for the regulated carriers to compete ef- fectively in certain areas and make it hard for government regula- tors to ignore the Internet. Finally, the carriers can only go so far in providing Internet services. Ultimately, the communication pathways must enter the userâs machine, pass through layers of software and end up in appli- cations programs. The computer industry, along with the many ven- dors of computer-related equipment, must play a role in determining how this aspect of the Internet will evolve. The nature of technologi-
22 ROBERT E. KAHN cal innovation almost guarantees that many new technological op- tions will continue to be generated from many different sources and make their appearance throughout the Internet. Thus, it appears that no single entity can possibly be in charge of the Internet. A key to the success of the Internet is to insure that the interested parties have a fair and equitable way of participating in its evolution, including participation in its also-evolving standards process. A proper role for governments would be to oversee this process to make sure that it remains fair and meets the wide spectrum of public needs. An international infrastructure like the Internet will ultimately require countries to set policy on many of the details that are now taken for granted. For example, Internet names and addresses may take on additional legal meanings in the various countries as they rely on the Internet to a greater degree. Trademark of Internet names and addresses is only one aspect of concern. Contracts of all sorts may have Internet names and addresses embedded within them. How can the countries have confidence in the use of such names and addresses for legal purposes without necessarily assuming responsibility for the day-to-day operation of this aspect of the system? Computer viruses know no national boundaries. If a major âinfectionâ should strike multiple countries, how will those countries work together to respond to such a situation? Finally, the ability to conduct network-based business between countries will require the resolution of many legal issues, including the formalization of legal contracts online and the ability to deal with associated customs and trade-related matters. At its core, the issue of online legal contracts seems to require the use of encryption technology, which has been perhaps the most closely held of all the network-oriented technologies. How can this kind of capa- bility be made available in the international arena in ways that are acceptable to national authorities? More generally, how can issues like those described above, which are likely to arise in the future, be effectively discussed and resolved? Various subsets of these kinds of problems have arisen in the context of other international public networks, including for tele- phones, and are thus neither unique nor entirely new. As the Internet continues to grow, many of the approaches developed for earlier technologies may apply to the Internet. Some combination of public- and private-sector involvement will probably be required to deal with these problems more generally.
EVOLUTION OF THE INTERNET 23 Governments have a fundamental role to play in the funding of advanced research and development that can push forward the fron- tiers of technology and knowledge. Often, this will involve the development and use of pilot projects to test new ideas in the real world. It also seems clear that governments must provide the neces- sary oversight to insure that the standards-setting process is equi- table. Governments must also take responsibility for helping to resolve problems that arise because of independent decisions made by multiple countries, for example in legal, security, or regulatory matters. In the case of U.S. infrastructure development, the govern- ment must provide leadership in many dimensions, including the removal of barriers where they inhibit progress; the insertion of legal, security, or regulatory mechanisms where the national interest so dictates; and the direct stimulation of public-interest sectors, for example in research, education, and certain network aspects of pub- lic health, safety, and universal access that require government as- sistance. Other nations also may find similar incentives for govern- ment involvement. Two final observations seem appropriate. First, it will be essen- tial to separate the process by which standards are selected for the Internet from the process by which the variety of possible options are generated. The current situation is almost ideal, since standards are selected by a process akin to ratification only after independent imple- mentation has produced the viable options. This separation needs to be maintained. Second, the most important use of the Internet, and indeed the NII, will be to allow individuals to communicate with each other and to rapidly access information. In many cases, this information will be the intellectual property of others. Every Internet user will also have the opportunity to become a potential provider of information ser- vices, thereby vastly increasing the amount of information available. How much of this information may be deemed valuable in a literary or business sense remains to be determined, but much of it may be important in other contexts. It is essential that we sensitize individu- als to the value of intellectual property and the need to protect it. This will have the side benefit of encouraging others to develop and make available intellectual property of their own. A combination of ethics, technology, and law are needed to ensure the effective devel- opment of this important aspect of the Internet.
24 ROBERT E. KAHN CONCLUSIONS Over a span of some 20 years, the role of the U.S. government in the evolution of the Internet has changed. While the federal govern- ment took the lead in virtually every aspect of Internet in the early days, it currently plays a more limited role. The government is now a major funder of network R&D and provides significant oversight of the evolution of the Internet. It provides direct support or even con- trol for several key aspects of the Internetâs operation, such as the assignment of unique names and addresses and the assurance of ad- equate backbone capability, although it may decide to relinquish some of these responsibilities in the future. It continues to stimulate the development of Internet architecture in healthy new directions. Although the role of the U.S. government in the Internet has been declining steadily for several years, particularly as private-sec- tor interest in the Internet has increased, there is a major continuing set of roles and responsibilities for government to undertake, both in the United States and around the world. Governments must be in- volved in decisions about how different countries cooperate on vari- ous aspects of the Internet and its use, and they must continue to oversee the networkâs evolution, both nationally and internationally. Other national governments may, but need not, assume the leadership role that the U.S. government has traditionally played in the United States. Without substantial U.S. involvement however, it is doubtful whether the NII will become a reality. And without government involvement on an international scale, it is unlikely that a global information infrastructure will emerge or that the Internet will con- tinue to evolve in a vital and dynamic way. Taking a long view, network and computer technologies are still in their infancy, and many of their current uses reflect past practices carried out more effectively in new environments. The real challenge will be for the public and private sectors to work together to harness the still-untapped potential of new and increasingly powerful tech- nologies in the network-based setting of the NII, and to nourish and incubate the powerful, even revolutionary, new ideas that are certain to surface in the future.