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 Promise of Infrastructure The board believes that the effective use of computer technology will increasingly require a networking infrastructure. We envision a nationwide computer communications capability that would enable any computer in the United States to communicate with any other computer easily, reliably, and over a broad range of speeds com- mensurate with individual application needs. This capability could accelerate the conversion of computer technology advances to prac- tical uses in businesses and homes, and it could help businesses and other organizations increase their productivity. The board has addressed some of these issues in reviewing federal proposals to unprove networking for U.S. researchers (CSTB 1988~. It sees in that narrow context powerful options for improving the way the community at large does its business. In advance of an effort to study the issues associated with that much larger goal, however, we review the motivation for our interest in improving networking on a national scale. INFO1~IATION NE:TWO1~S Almost 50 years elapsed between the invention of the telephone in 1876 and completion of the national toll network. The first na- tionwide plan for providing good random access service was put forth in 1925. A long tune, perhaps, but today we take voice telephone 14
15 service for granted. We can call anyone, anywhere. We know how to operate virtually any telephone anywhere. If we do not know a per- son's number, we can look it up in a telephone book or call directory assistance. Once a can is connected, we can talk as fast as we are able, and we can use any language we wish. The telephone network han- dles it Al with ease. Our commerce, our pleasures, and our everyday life have come to depend on the existence of this richly connected, ubiquitously available voice communication highway system. By contrast, ahnost none of the characteristics just cited applies to the communication of information between computers. The com- puter networks of today, and often individual computers, are islands unto themselves. They are not interconnected, and it is not possible to send information between any two systems. The members of the Computer Science and Technology Board, for example, are members of leading institutions ~ computer technology in our nation, yet we are often unable to send electronic mail among ourselves with- out great difficulty and without using pathways and circumventions unavailable to the general public. To understand thm, consider that, in order to make a voice telephone connection, in principle it ~ necessary only to connect two electrical wires. To allow computers to communicate is much more difficult because, in addition to physical connectivity, there must also be logical connectivity. To explain: computers transmit and receive blocks of bits, which must be packaged in particular formats. For example, each package must contain the address of the intended recipient, the sender's or the return address, the data being communicated, so-called check bits intended for error control (one erroneous bit can ruin a message with millions of bits), as well as other routing and control information. Designers of communicating systems must agree on exactly how these packages of bits are to be constructed. In addition, every user must agree on the sequence of steps required to establish a connection, and on what procedures must be followed in the event of a transmission error or under any of a number of other exceptional conditions. The collection of formats for packaging data and the rules that govern the logical flow of data transmission are caDed data commu- nication protocols. While there has been a significant international effort to standardize on several protocols, the attendant development of consensus has taken too long relative to the pace of computer tech- nology, and many small networks have been constructed using many different protocols. In effect, these different networks do not speak
16 the same language and are unable to communicate with one another, except perhaps with a great deal of effort. In computer networking, just as in transportation networks, the speed at which individual users can operate is an important consid- eration. For decades the nation lived with and became accustomed to a teletype network that transported data at about 150 bits per second, a rate roughly equivalent to fast human typing. In the 1960s, computer modems were introduced to connect computers to the voice telephone network. Starting at rates of 110 bits per second, these devices have evolved so that today most data traffic ~ transmitted over voice facilities at 1,200 bits per second. At this speed, a page of text appears on a display screen in about ~ seconds. While this speed may be appropriate for displaying messages, there are many appli- cations, especially in computer-to-computer communication, which would thrive on greater, even much greater, speeds.* Imagine trying to skan through a book to find a particular section if it took 8 seconds to turn every page or trying to access remote supercomputers that consume and generate data at millions of bits per second. The extensive wiring of the nation with optical fibers by the com- mon carriers represents an unport ant national facility for high-speed computer communication. The latest fiber systems have data rates of 1.7 billion bits per second, equivalent to some 50,000 simultaneous voice telephone cans, on each ha~r-thin fiber. Moreover, the progress in lightwave transmission during the last decade has been such that the transmission capability of fibers has doubled each year. We ex- pect rapid progress to continue for at least another 5 to 10 years. We also welcome the growth in digital communications services and, in particular, the movement toward voice-data integration, network standards such as the Open Systems Interconnection (OSI) model, and higher speed services available to individuals (e.g., through inte- grated services digital networks (ISDN) being introduced gradually by common carriers). But the current and anticipated situation for computers ~ very much as if we had built a superhighway system spanning the nation without on and off ramps, without a connecting *A high-resolution screen in today's workstations often displays a mixture of text and graphics comprising several million bits of information. A typical program used even for such mundane purposes as word processing is of roughly the same size, measured in bits. Engineering drawings and photographic images are yet other examples of information requiring millions of bits to describe. We cannot transmit quickly or economically these kinds of information over current long-distance data networks.
17 network of local access roads, and without common understanding of vehicle speeds, widths, and loads. A national information networking capability could build on evolving digital communication facilities. It could extend access to a number of information and net~vork-based services now only available to relatively affluent computer users and make possible new public- access services that can only be sustained if done on a sufficiently large scale or that require state-of-the-art networking technologies (see Figure 1~. If the history of the highway and telephone systems is any indication of what the future may hold, a new infrastructure based on computer networking might also result in an array of new information industries and businesses that we cannot even foresee at this tune. Nevertheless, in what follows, we describe some of the uses we envision from our current vantage point. USES OF INFORMATION NETWORKS ON A NATIONAL SCALE The potential uses of expanded, nationwide computer network- ing are suggested by existing systems as well as by our understanding of emerging technologies. Information and information-related ser- vices are already bought, sold, or otherwise transacted within and between enterprises. For example, many traditional service indus- tries (e.g., finance, insurance, accounting, and law) depend heavily on information as a product or a component, while newer services (e.g., electronic mail, electronic access to bibliographic data bases, and large-scale systems design and integration) have emerged as a result of developments in computer technology and its uses. Existing networks also provide glimpses of the potential of net- working to enhance productivity in the manufacturing sector. Com- puter networks can facilitate collaboration among Aspersed design teams and enhance interaction among distributed design, manu- facturing, and marketing personnel. Such activities have begun to emerge at the most progressive companies, and we expect their ef- fectiveness to grow with expert system and multiprocessors. Com- merce can be made more efficient by the electronic exchange of orders, invoices, and payments, in lieu of much slower exchanges by conventional mail. The automation of ordering, invoicing, and so on, is caller! electronic data interchange (EDI). Today, some firms use their own networks for EDI, while others rely on shared networks and
18 fry A! 'if I X 0 o A\ A W~ ~ ~ ~. E oil / ~ 'a -m r. ;~ ,_ _, _~ ~ ·m ._ 4. O ~ o ._ ~ be ·= To Ll . be _- 0 ,Y ret ~ ~0 a, _ ._ ~ on 0 0 C) to . En L. O · 3 0 AL ~ o d ~ `6 ~ ~ _ ~ ~O4 m P. ~ a' cot o ~.t, so ~ ._ ~ 4_ 3 os o O · _ 0 ~ ~ o _ _ C,) 0 ~ 0 d 0 ·o A ~ at =.0·> ~ ~ ~0
19 third-party computer services. In either case, the computer networks they use for such transactions are limited in scope and service. Networking on a national scale would support (economically) the establishment of a variety of computer-based, public access informa- tion resources, such as a national digital library which might make it possible for anyone, anywhere in the nation, to access and read any book, report, magazine, or newspaper (at a nominal cost). In the research community, we have already seen how achieving the bene- fits of supercomputers is closely intertwined with networking, which allows remote access to these scarce and costly facilities as well as collaboration among dispersed researchers (CSTB 1988~. Realizing the potential for new, information resources would require not only advances in networking, but also advances in storage, retrieval, cod- ing and classification systems (to facilitate information access and maximize its usefuIness), and user interfaces. Other countries, smaller in scale and slower to deregulate their communications industries, have established more or less nationwide public data networks run by governmental postal, telephone, and telegraph authorities (PTTs). Those networks have offered Innited protocol, equipment, and speed support, but in some cases PTTs have begun to explore the potential for nationwide information ser- vices. Perhaps the most frequently-cited example is the videotext service In France that combines Minite! terminals (provided by the government) with a public data network. Through Minitels, many information-based services are available through a uniform mode of access to consumers: telephone directories, travel and entertainment schedules, shopping, and more, but only a few have large customer followings. The Minite} experience underscores the challenges of implementing information services on the large scale and in the free- enterprise environment that characterize the United States, let alone for providing the richer service offerings that emerging technologies can make possible. In this country, the greatest progress toward nationwide, inter- organizational networking service has been achieved in the research community. For example, the government launched Arpanet through DARPA in 1969. It has been a mode! for subsequent public data net- works, including those offered commercially; and in the research world, it has been complemented and augmented by such general- purpose networks as Bitnet and multiple special-purpose networks (e.g., the Space Physics Analysis Network). The fragmentation of research networking, the low quality of many research networks, and
20 the benefits of nationwide interconnectivity for researchers, have led to proposals for a national research network with widespread access sibility, high speeds, and a variety of associated information services (OSTP 1987~. But even this specialized project raises many ques tions about costs, financing, the role of computer, communications, and other companies, the role of the government, and so on, as wed as appropriate technology (CSTB 1988~. Note that the goad and benefits of a national research network can be met by a loose feder- ation of smaller, private and public networks; ~national" should not necessarily be taken to mean monolithic or even government-owned. Serving the economy as a whole, public data networking (through value-added networked and such network-based services as ED! and electronic mail are commercially available in this country. The de- velopment of the markets for these services has been slower then projected and, for much of this decade, unprofitable. The growth of these markets is tied to growing recognition of the benefits of networking computers, growing comfort among a wide range of peo- ple with the use of computers, as well ~ the spread of computer equipment and the development of applications that combine data processing with communications. It is also expected that emerging standard (e.g., standards for ED! document formats, for voice-data integration, for logical connectivity of people engaged in applications, and for electronic mail system interoperability) wiD contribute to the demand for network-based services. The board recognizes that achieving its vision of nationwide computer networking with greater speed, logical connectivity, and accessibility as well as a richer menu of services than now available raises many questions including the following: . What are the principal technical obstacles, and what would be involved in overcoming them? Does nationwide service require a single physical network? ~ 18 industry likely to supply the necessary features and services on its own, and if 80, in what time frame? ~ What would this national capability cost, and how should it be paid for? ~ What social, economic, and legal side-effects and adjustments might be involved? How can networking be made easy to use for all prospective users without compromising the privacy and security of users and their applications?
21 . What would be the most effective roles of government, indus- try, and other entities? None of the above problems is insurmountable, but, as the above list suggests, planning and management as well as technology will be important in enhancing computer networking on a national scale.
