IV. THE INTERNET

DEFINING THE INTERNET

The Internet was chosen for study because of its current social and economic impact rather than because of any particular technological breakthrough or historical precedent. It is thus appropriate to begin analysis of its origins by characterizing the Internet as it now exists-the technological, organizational, and institutional features that account for the impact we observe. Once this is accomplished, we can decompose the innovation into its technological components and trace selected underlying technologies to their origins. Because the Internet is not embodied in physical form, patented, or copyrighted, the definition and decomposition tasks are themselves challenges.

Reflecting the fact that the Internet is a capability that has no single organizational or corporate form, it can be characterized in various ways: as a network of networks based on the TCP/IP protocols, a community of people who use and develop those networks, and a collection of resources that can be reached from those networks (Krol and Hoffman, 1993). According to Vinton Cerf, one of the Internet's pioneers,

The name "Internet" refers to the global seamless interconnection of networks made possible by the protocols devised in the 1970s through ARPA-sponsored research-the Internet protocols, still in use today. ... The Internet now encompasses an estimated 50,000 networks worldwide, about half of which are in the United States. There are over 5 million computers permanently attached to the Internet, plus at least that many portable and desktop systems which are only intermittently online. ... Traffic rates measured in the recently "retired" NSFNET backbone approached 20 trillion bytes per month and were growing at a 100% annual rate. (Vinton G. Cerf, http://www.cs.washington.edu/homes/lazowska/cra/networks.html)

A recent report on the Internet and its successors adds that about half the networks that comprise the Internet are commercial and the proportion is growing, and that about a third of the remaining, noncommercial networks are associated with research or educational institutions. Approximately 150 countries are connected in one way or another to the Internet (National Research Council, 1994: 21).

Brian Kahin's definition reflects the Internet's functional yet disembodied reality:

It encompasses the set of interconnected and interoperating networks which make use of the internet protocol (IP) and a common addressing scheme; the Internet is international in scope and encompasses some 5000 autonomous local area networks as well as dozens of autonomous higher-level networks (midlevel networks as well as national or international "peer" networks). An uppercase I is used to distinguish the one global internet from the hundreds of thousands of lesser internetworks, which may or may not connect with the Internet. However, it may convey the false impression that the Internet is a service or system or that it has some institutional or corporate embodiment.

As a practical matter, the Internet is not delimited by the internet protocol but by the broader notion of interoperability. Parts of the Internet support multiple protocols, including OSI CNLS, the connectionless packet-switching standard adopted by the International Standards Organization (ISO). So the Internet should probably be defined as wherever one can get from the core Internet with full bilateral Internet functionality (mail, file transfer, remote log in). Thus, networks with simple mail gateways into the Internet are not normally considered part of the Internet. (Kahin, 1993: 7)

On the basis of these definitions, the Internet can be characterized as a system that embodies the following technological components:

• A set of interconnected communication networks (internetworks).
• Common protocols governing addressing, routing, and error-checking of information packets.
• Packet switching: information on the network is formed into packets from multiple, simultaneous users, then sent and reassembled at destinations using sequencing and routing information contained within the packets.
• Switches, or routers, at each node of each network that control the routing of packets.

Internetworks

The various levels of networks, their physical characteristics, the functions they perform, and the institutions involved can be best understood within a four-level conceptual scheme introduced by the Office of Technology Assessment in its 1989 background paper High Performance Computing and Networking for Science (OTA, 1989).

Level 1: The physical network infrastructure consists of the fibers, wires, cables, and radio and microwave links that carry digital signals. In the United States, this infrastructure is owned by common carriers (e.g., AT&T, MCI, Sprint).

Level 2.: The user-defined subnet's function is to route messages among users, deal with congestion, and interconnect among different networks. This level of services is provided by the regional, state, local, commercial, and government agency networks.

Level 3: The network services level supplies end-to-end connection, electronic mail (e-mail), file transfer, remote login, network monitoring, and network security.

Level 4: The value-added superstructure consists of links of the network to "specialized computers and software, library catalogs and publication databases, archives of research data, conferencing systems, and electronic bulletin boards and publishing services that provide access to colleagues in the United States and abroad" (OTA, 1989).

Protocols

The suite of application and transport protocols used on the Internet has been labeled TCP/IP (Transmission Control Protocol/Internet Protocol). Basically, IP is the addressing system and TCP controls how packets are assembled at the origin, routed, tracked, and reassembled at the destination. Also included in the suite are FTP (File Transfer Protocol), telnet (remote connection protocol), SMTP (Simple Mail Transfer Protocol), and UDP (User Datagram Protocol). FTP and the others listed add functionality to the basic TCP/IP software but are not intrinsic to it.

The basic TCP/IP protocol suite was designed in 1973 by Robert Kahn (ARPA at the time) and Vinton Cerf (Stanford at the time) for the ARPANET[61] and eventually became the standard protocol for the Internet. At the time, it was the only protocol that dealt explicitly with the internetworking of packet-switched networks; without it, incompatible networks such as the separate radio, satellite, and ground-based networks used by the military could not be connected (Kahn, 1995: 17).

Packet Switching

Packet switching is

a data transmission technique whereby user information is segmented and routed in discrete data envelopes called packets, each with its own appended control information for routing, sequencing and error checking; allows a communications channel to be shared by many users, each using the circuit only for the time required to transmit a single packet. (Data Communications, 1988)

The idea that makes packet switching work is that messages occupy a communication link only while data is being sent, rather than on a permanent basis, as in the case of circuit switching. This approach represented a fundamental paradigm shift from the prevailing, circuit-switched communication networks. The idea was put forth by Leonard Kleinrock while he was a Ph.D. student at MIT in 1961.[62] At the same time (1960-62), Paul Baran of the RAND Corporation (supported by the Air Force) was applying the concept of what he called "message blocks" to distributed networks as a solution to the problem of communication network survivability during war. The motive behind the idea was to design communication systems that were more "survivable" during war; if part of the network were destroyed, messages could still be sent via surviving nodes. Independently, Donald Davies of the National Physical Laboratory in the United Kingdom was working on similar problems and coined the term packet switching in 1966. Building a functioning packet-switched network was a different matter; the first such network was actually built and operated with ARPA[63] funding in 1969. The first node was constructed at UCLA in 1969 under Leonard Kleinrock's direction; three other nodes followed quickly at the University of Utah, the Stanford Research Institute (now SRI International), and the University of California at Santa Barbara (Norberg and O'Neill, 1992: 242-249). This was the beginning of the ARPANET.

Routers

These are the combination of hardware and software that exist at nodes in packet-switched networks and serve as “intelligent switches” for packets. “The intelligent switch collects a sequence of bytes from an incoming device’s byte stream to form a data block, adds a header to this block (containing, for example, the destination and source address for this block), adds a tail to this block (containing, for example, error control information), and calls the whole thing a packet. These packets are then launched into the network of intelligent switches whose further job it is to route these packets through the network in an adaptive, dynamic fashion so as to deliver them expeditiously to the appropriate destination” (Kleinrock, 1993: 175).

Browser Software

The Internet requires all the above elements, but its recent and continuing explosive growth has been strongly influenced by the existence of user-friendly, powerful “browsers” (sometimes referred to as "user interfaces"), such as Mosaic and Netscape, and the proprietary software incorporated in subscription services such as America Online and CompuServe. In the absence of this category of software, it is highly unlikely that the Internet would have expanded so rapidly or encompass such a variety of users and sources of information. Thus browsers are an additional technological element that enables the Internet to be used as a powerful information search device.

What evolved into the World Wide Web (WWW) was created in 1988 by Tim Berners-Lee, a physicist at Europe’s high-energy physics facility, CERN. Released in 1992, it enabled users to embed Internet addresses in their documents. Readers could simply click on these references to connect to the reference location itself. Although originally created to assist the community of high-energy physicists, the WWW became an essential ingredient in the search capability of browsers.

Soon after its release, the WWW came to the attention of a programming team at the National Center for Supercomputing Applications (NCSA) at the University of Illinois. This team developed Mosaic, a graphical browser for the Web that enabled users to provide images, sound, video clips, and multifont text within a hypertext system. Subsequently, numerous commercial versions of Web browsers, such as Netscape, were introduced and have generated substantial fortunes for their creators (Vinton G. Cerf, http://www.cs.washington.edu/homes/lazowska/cra/networks.html).

It would be a mistake to define the Internet solely in technological or functional terms because of the importance of organizational and managerial innovations to its evolution. Indeed, our initial conception of the Internet in technical terms (as would be appropriate to the two other innovations studied in the first year) was quickly superseded as a result of our initial interviews with SRI International researchers and NSF program managers, and an initial reading of the literature. As our respondents at SRI put it,

In addition to technologies, you should include organizational innovations, especially NSF's role in insisting on the formation of consortia of universities and companies. NSF came up with ways to get people to work together. This had a lot to do with getting protocols accepted. NSF, NASA, and ARPA (and perhaps DOE) worked very closely together in providing seed money. NSF was the partner representing universities. NSF played a lead role in creating and encouraging consortia. (SRI interview with D. Nielson and E. Feinler, January 29, 1996)

Our subsequent investigations not only confirmed this point, but provided additional examples of NSF's managerial leadership. Thus to our list of defining features of the Internet we add organizational and managerial factors.

Organizational and Managerial Factors

There have been innovative organizational, managerial, and administrative decisions and procedures that were highly influential in the evolution of the Internet. Institutional roles have changed dramatically over time, beginning with exclusive control by ARPA over ARPANET, the Internet's precursor, spreading to involvement of several other agencies, including NSF, in the late 1970s and early 1980s, to a major NSF role (as far as government was concerned) but in partnership with private firms from the mid-1980s to mid-1990s, to the current situation in which there is no direct government support except for some "connections" programs.

In the 1960s, "networking" on university campuses consisted mostly of linkages between dumb terminals and timeshare hosts. The exception was ARPANET, which was beginning to link compuses with ARPA computer research contracts on a host-to-host basis. By the early 1980s, other federal agencies, such as DOE and NASA, had created their own research networks (e.g., DOE's HEPnet and MFENET, NSF's CSNET), but connections among them were difficult at best. Access to these networks was restricted to computer science departments and researchers working for the federal agencies involved. ARPANET already connected a large number of research computers, but the idea of using the supercomputers of the old CSNET as the "backbone" for the new NSFNET is attributed to NSF’s Dennis Jennings in 1985, and it was Jennings who insisted on the TCP/IP protocol for the new backbone, a fateful decision (Merit, 1995: 19). Under the leadership of NSF program managers Jennings, Steve Wolff, and Jane Caviness, NSFNET evolved to a three-tiered architecture: backbone, regional networks, and campus networks. The backbone was managed by Merit, a consortium of IBM, MCI, and the University of Michigan; regional networks were managed independently under a variety of arrangements (Mandelbaum and Mandelbaum, 1993: 81), and the campus networks were managed locally. Merit subcontracted support of NSFNET to the newly formed nonprofit Advanced Network & Services, Inc. (ANS) in 1990. Beginning in 1987, NSF issued a series of solicitations that increased the performance of NSFNET and worked toward a system in which both the regional networks and the backbone would no longer require government support. In April 1995, NSFNET was decommissioned, and operation of what we know as the Internet switched from public to private hands-with hardly anyone noticing. As Vinton Cerf describes the transition,

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 (ARPANET, SATNET, PRNET and NSFNET) are now gone-but the Internet thrives! (Aboba, 1993)

This completes our detailed description of the Internet, broken into its technical and administrative elements to facilitate our analysis of NSF's role in its evolution.

AN INTERNET CHRONOLOGY

Because of its significance, the Internet has been the subject of numerous historical accounts, some scholarly and some popular. We certainly do not have to construct such a history ourselves, since many exist. Fortunately, we can draw on these histories and focus our resources on our primary research question: How and at what points has the National Science Foundation contributed to the evolution of the Internet? To lay the groundwork, the following is a brief chronology of major milestones in the development of the Internet: