Click for next page ( 31


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 30
Software Growing at a rate of about 25 percent a year, estimated global revenues from sales of software exceeded $65 billion in 1989.1 Moreover, when the value of software developed within businesses and other organizations is taken into account, the size of the market for software is much larger. Measured by salaries and other costs, the value of internally generated software in the United States may range between $150 billion and $200 billion.2 These figures are compelling evidence of the attractiveness of the software market, which is dominated by U.S. fobs. They also attest to the increasingly important role that programs and their applications play in enhancing invest- ments in hardware. Moreover, the growing emphasis on software development spurs the innovation of new equipment, which in turn creates ideas for new applications that require new software. "As software gets developed," explained Alan Perlis of Yale University, "it exposes opportunities for the application and creation of hardware. The two work together, and there is no permanent boundary between them. There is a continuous shift, one way or another." Thus, for example, the emergence of new processing architectures, notably parallel processing, gives rise to a need for a whole new body of software which may in turn enhance the develop- ment of new hardware. The current situation has changed dramatically in less than 15 years. Prior to the introduction of the personal computer, software functioned somewhat as a "loss leader," an often-free inducement for buying and continuing to buy a particular vendor's computer equipment. Software development was part and parcel of hardware development, and computer manufacturers were nearly the 30

OCR for page 30
SOFTWARE 31 sole source of programs, which were written to run on their machines only. The few independent software firms that did exist wrote programs tailored to indi- vidual vendor's machines. Sales of about a thousand copies were the bench- mark for a highly successful software package. Starting with Apple Computer's popularization of the personal computer, the size of the software market grew immensely, spawning small and large firms that derive all or most of their revenues from this market. Today a million copies of popular software packages may be sold. Moreover, increasing but still modest standardization has made it somewhat easier for developers to adapt programs for use on the computers of different manufacturers. In addition, even the largest computer companies recognize that they have neither the financial resources nor the technical staff necessary to provide the full array of software support that potential buyers of their machines now demand while larger buy- ers of software, with similar resource limitations, look to the software industry to meet their needs affordable. Out of these seeds grew the burgeoning software industry. Some U.S. soft- ware fops are now very large, the biggest reporting annual revenues of about $1 billion, but most are quite small, sometimes consisting of not much more than one or two people "with a personal computer and an idea." This diverse collection of U.S. firms accounts for more than 60 percent of world sales of software.3 At present the U.S. industry enjoys a commanding lead in the world market, but hindsight suggests that no lead in markets for computers and com- puter-related parts is secure. "Assuming some of the dire predictions . . . come true," said Samuel H. Fuller, vice president for research at the Digital Equipment Corporation, "then a large part of our hardware twill be] . . . manufactured and eventually designed outside the United States. Is there the kind of infrastructure and support In the other industries of the computer sector that they could remain, in fact, healthy industries in their own right? Can software exist as an industry without the supporting underlying hardware?" The answers to these questions are not clear- cut, but Fuller suggested addressing potential problems before they become real ones. Software "is an area where we have to learn how to remain strong, rather than take it for granted," he said. SOFTWARE: A HIGHLY UNUSUAL PRODUCT Arguably one of the most important products of the age of high technology, software has properties that make it distinct from all other products. The largest and most sophisticated sets of programs, such as those written for air traffic control systems or early-warning defense systems, are among the most complex of human creations, rivaling the crowning physical achievements of modern engineering. Yet software development has features that make it more akin to an art or a craft than to a high-technology enterprise. It is an expensive, slow,

OCR for page 30
32 KEEPING THE U.S. COMPUTER INDUSTRY COMPETITIVE and labor-intensive process that has undergone only incremental improvements in productivity.4 Another special characteristic of software is that it typically does not remain constant. Observed Perlis, "The thing that has been missed by software engi- neering to date, to everyone's great peril, is the concept of evolution that soft- ware, as soon as it comes to exist, must change. The most successful software is that which is unstable and incomplete." Perlis explained that "unlike hard- ware, software gives the illusion that the costs of change are negligible since they apparently involve only scribblings. This is not so. Software affects its specification so that both undergo continuous change." Other major properties that distinguish software, according to Perlis, are net- working-toward the end of direct communication between programs, and automation since "it must be the case that most of our programs run essential- ly independently of us." Finally, Perlis explained that "in the computer, pro- grams are in a sense live.... They can do things we fold extremely difficult to do-that is, to keep informed in a broad way about what is going on in the world." Perhaps the most complex pieces of software entail integrating isolated col- lections of computer, communication, and other types of equipment into coher- ent information-processing systems. Systems-integration software, the means of accomplishing this complex linkage, often must be customized to meet the diverse requirements of customers, explained Laszlo A. Belady, vice president and program director for software research at Microelectronics and Computer Technology Corporation (MCC). Because of this customization process, soft- ware development can be considered as much a service activity as a manufac- turing process. Software is itself increasingly valuable in the production and modification of software. Researchers throughout the world are developing productivity- enhancing tools and methods for specifying, writing, editing, compiling, testing, and verifying programs. Although many are the proprietary possessions of indi- vidual companies, the tools already in hand have helped shorten the software development cycle, particularly for easily specified, straightforward applica- tions. However, continued leadership in software calls for greater investments in developing and using software development tools. Noted Lawrence Tester of Apple Computer, "In order to compete with, say, the Philippines where the labor costs are lower for programmers, we need to lower our overall cost of produc- ing software by improving our tools." In the software industry, low levels of productivity growth are a shared short- coming, besetting firms in the United States, Japan, Western Europe, and every- where else. In the United States, the most visible efforts to enhance productivi- ty in software development, including the software research programs at MCC and the Software Productivity Consortium (SPC), were encouraged by frus- trated users, notably defense contractors interested in reducing programming

OCR for page 30
SOFTWARE 33 costs. The software industry has not pursued a coherent strategy to improve software development, not the least because of the implicit costs and uncertainties. TOWARD A BIFURCATED MARKET Current trends in particular, the emergence of large-volume markets for packaged software and the growing demand for customized software-are stim- ulating a division within the industry, points out Laszlo Belady of MCC. One segment, he said, will concentrate on software packages that, although varying in size and complexity, can serve many different customers without major cus- tomized alterations. "These packages can be produced anywhere," said Belady, suggesting that U.S. hums serving this market will be most vulnerable to foreign competition because "everybody has the same chance." In Belady's second category, systems integration, competition will be deter- mined by the ability to manage complexity, to develop applications tailored to the idiosyncrasies of individual enterprises: "What you have to do in order to make this complex integrated application work is to provide the glue, that is, additional software, which does the traffic control and holds the pieces togeth- er." This category, according to Belady, is where the greatest business opportu- nities may lie and where U.S. firms may have an inherent advantage. Successful software of this second type can only be developed through extensive cooperation between the customer and the contractor, Belady empha- sized, and "it is impossible to come up with a huge computerized integrated application for an enterprise where the vendor comes and does everything . . . it cannot be done without team work." For members of such teams, "it is not enough to teach programming," he said, "but you also have to give real-life experience, even at school, and teach people how to work together and not just to excel individually." Belady explained that if we wish to master the necessary complexity of this second software category, we must exploit the potential of computers them- selves in retraining individuals. He described retraining as "an incredible opportunity, an incredible benefit to the country." Belady stated that retraining could be invaluable in making this country more flexible and able to cope with the many new challenges that the "accelerating change of industry and society" confronts us with. EDUCATION FOR BETTER SOFTWARE DEVELOPMENT AND USE The link between education and the ability to stay competitive was made succinctly by Tester: "If other countries have better-educated populations in computer science and software engineering, then we will fall behind." Many colloquium participants commented on the promise for improved education and training as means to achieve better software and better software development.

OCR for page 30
34 KEEPING THE U.S. COMPUTER INDUSTRY COMPETITIVE Colloquium participants cited university computer science programs as a particularly valuable asset to the software industry, because they provide needed personnel. "At times, people will argue that because of our education and because of our innovative spirit," observed Fuller, "we [the United States] will continue to hold a commanding lead in the area of software." The proportion of American-born students enrolled in these programs has dropped substantially, however. "Whereas it is bothersome that half of our graduate students are from foreign countries," said James H. Morris of Carnegie Mellon University, "we should try to make that a competitive advantage and give them every opportuni- ty and encouragement to stay, both at the graduate and undergraduate level." Getting the most out of software may, as Perlis suggested, entail developing systems that "get the human being out of the loop," but achieving that end requires human talent and skill. All colloquium participants had suggestions for improving educational programs in computer science and software engineering. Tester, for example, pointed to the need for a practical perspective, commenting that "in terms of computer science and software education, what I would like to see coming from the colleges is people with more practical experience . . . fincluding experience doing] major projects on teams to revise other people's software, which is what one ends up doing in industry quite a lot." The value of increasing the exposure of faculty and graduate students to industrial software problems and development conditions has been noted elsewhere by the Computer Science and Technology Board.s A recurring theme was the importance of interdisciplinary training that goes beyond software and hardware issues per se. Tester related such training to maintaining a specific competitive advantage: "One thing that I think the United States can maintain a lead in is human interface design, which makes application software distinctive, but to do that, our students have to be very broad in their education. They cannot learn only technology; they have to learn something about psychology, something about art, and learning to work in teams with people of other disciplines.... We can stay ahead, particularly in the Far East, . . . by maintaining these cross-disciplinary development teams." But participants' concerns about education extended to the general popula- tion. The health of the entire computer sector and the development of the tech- nology will depend to a great degree on how society chooses to use computers and extend their applications. "The important thing to remember," said Yale's Perlis, "is that software is intended, as is the computer, . . . to make life more imaginative for all of us." He added that, because of the technology's influence, society will be changing continually. Therefore, Perlis and others said, computing must be integrated into all areas of study. From growing familiarity with the technology among broad segments of the population, ideas for new applications will emerge, fostering the evolu- tion of the technology and driving the growth of the software industry and oth- ers. "The effective use of computing," said Abraham Peled, vice president of

OCR for page 30
SOFTWARE 35 systems research and director of computer sciences at IBM's T. J. Watson Research Center, "will be the key to the industrial competitiveness of the coun- try as a whole, as products are conceived and developed inside a computer instead of the laboratory, and integrated directly with manufacturing [and] with the field force." Education is an immediate concern. "If we see it as long-term, and therefore not requiring immediate attention, we will never get around to it," emphasized Tesler, "and I think that is what we have been doing in this country for the last 20 years." FOREIGN CHALLENGE Some erosion of the nation's lead in software markets is inevitable, but if limited, it could have positive effects, according to Apple Computer's Tesler. "Some degree of decline is healthy for the world economy," he said, and it will "stimulate more attention by U.S. software companies to international markets." Already, noted Tesler, "For routine types of software creation, the sorts of things that run businesses, . . . people are beginning to find that they can buy software development services from other countries ten times more cheaply than they can get them here." Several U.S. feds have already responded to this reality by establishing software development centers in such countries as Singapore and Ireland, where they benefit from relatively low labor costs. Domestic compa- nies also "import" temporary foreign workers to write code for basic, or standard, software packages. India, for example, advocates exporting its pro- grammers to work on international software conversion projects. How much of this activity takes place is not known, but it does involve leading U.S. compa- nies. Foreign production of packaged software is also growing. Tesler pointed out, "Some of the more innovative packages in the personal computer market have come from other countries, generally in Europe, say, France in particular. But as I have traveled around the world, I have seen interesting software in other places." To enhance their competitive prospects, Japan and Western Europe have con- cen~ated on developing productivity-enhancing tools and techniques, as well as methods for verifying the reliability of programs. To remedy gaps in expertise and to sidestep licensing restrictions, Japanese firms have also established soft- ware research laboratories in the United States. Other countries that have made their domestic software industries economic priorities, such as the People's Republic of China, India, Malaysia, and Taiwan, are focusing, for now, on large-volume reproduction of basic software rather than on innovation. Some Grins in developing countries, capitalizing on the low cost of replicating soft- ware and weak international protection for intellectual property, market imita

OCR for page 30
36 KEEPING THE U.S. COMPUTER INDUSTRY COMPETITIVE live packages.6 Such copied products undercut sales of legitimate products and drain resources for legal action for copyright infringement. The impacts of Japan's most visible efforts to achieve parity in software mar- kets are uncertain. Colloquium participants saw neither the Japanese Fifth- Generation Computer project nor The Real-Time Operating System Nucleus (TRON) project as yielding breakthroughs that could threaten the U.S. software lead. They were more cautious in their appraisal of a cooperative effort involv- ing government and industry to develop Software Industrialized Generator and Maintenance Aids (the SIGMA project). SIGMA "is a software production and industrialization system developed for the purpose of improving the fundamen- tal environment for software development in Japan,"7 building on AT&T's UNIX operating system. UNIX, developed by AT&T, is one of the leading can- didates for a standardized computer operating system for the entire industry. The some 40 Japanese companies participating in the SIGMA project have all agreed to use a specific version of UNIX. With a standard operating system, according to one line of reasoning behind SIGMA, Japan could be able to address deficiencies in software development. "The openness of UNIX," Morris of Carnegie Mellon warned, "makes it an ideal place for the Japanese to enter the U.S. software market. Look out." While ongoing developments clearly indicate a Japanese push to penetrate U.S.-dominated software markets, these efforts may not pose the most formidable challenge to U.S. firms. "The real software competitor is Europe, not Japan," MCC's Belady maintained. "'Europe 1992' will reinforce the suc- cesses of [the cooperative research program] ESPRIT in the area of standardiza- tion and formal methods 'creeping' into industrial applications."8 The European Community has also emphasized development of so-called formal methods for software engineering. If successful, formal methods could make some software development more systematic and less craftlike. Moreover, formal methods can also enhance software quality by giving more assurance that software does not contain errors and will perform as specified. Foreign competition has been constrained by U.S. dominance in computers, which has fed U.S. prominence in software. Although machine compatibility is increasing, getting the software developed for one vendor to work on the com- puters of another is far from effortless, and users are reluctant to switch to new hardware vendors and an associated new set of software vendors. Increasing standardization and, as has been the rule for three decades, continuing declines in the cost of hardware for a given level of performance could eliminate this barrier and open the door for increased foreign competition in software markets. Under these circumstances, it is imperative that U.S. software developers under- stand before their competition how to program computing structures that will emerge to achieve higher performance (e.g., multicomputers-distributed pro- cessing-and multiprocessors).

OCR for page 30
SOFTWARE 37 STANDARDIZATION Software research in Europe and Japan has a common thread, a push toward standardization. Underlying this push is the undisputed assumption that stan- dardization will improve productivity, lower the cost of software development, and increase competition. Virtually all software firms are advocates of these ends, but they disagree on whether standardization is the best means to achieve them. Discussions among colloquium participants were representative of this disagreement. Comments by Apple's Tesler characterize one view of the issue. "Any stan- dard that has to do with data communications with exchange of data between computers is absolutely critical," he said. He maintained that to standardize beyond that point would encourage competition from imitator feds and slow the rate of innovation in software and hardware. "Standards and what goes on inside a computer are another matter," Tesler said. "Standardizing processor designs by standardizing operating systems, my company feels, twill make the] United States very vulnerable to competition from abroad.... Japan and [South] Korea do very well when they can find something that has been standardized, or [they] create a standard and then lever- age that to reduce costs, reduce prices, and gain market share and take the mar- ket away." Tesler continued, "One thing that we [at Apple] have done is very carefully protect the insides of our computer, both the operating systems software and the hardware designs, so that they cannot be copied. We are very bothered by the fact that there are moves in the industry to try to standardize on those sorts of things, because we think . . . that is basically giving away the show to the com- petition from the Far East." Others in the industry are not bothered by the movement for a standardized operating system. As MCC's Belady noted, "Many people believe that stan- dardization is necessary for progress." According to this view, standards estab- lish a base level of conformance for well~eveloped technologies, freeing the industry to concentrate on areas where advances are likely to have a greater impact on the capabilities and applications of computers. Raj Reddy, professor of computer science at Carnegie Mellon University, argued that failure to achieve some standardization of operating systems would be counterproductive, diluting the efforts of programmers, who are already in short supply, and draining financial resources. "I think we should be standard- izing routine things, including operating systems, and trying to use our creativi- ty at higher levels," Reddy said. Preventing the adoption of standards is not the answer to maintaining com- petitive U.S. software and hardware industries, he maintained. If, in the future, foreign firms can produce powerful yet inexpensive workstations and U.S.

OCR for page 30
38 KEEPING THE U.S. COMPUTER INDUSTRY COMPETITIVE companies cannot, Reddy argued, the disparity will stem from a problem that is more fundamental than standardization. The implications are similar for firms that develop software only. By foster- ing high levels of compatibility among the computers of different vendors, stan- dardization would greatly expand software markets, uniting a fragmented cus- tomer base and eliminating the need to write unique code for each of the many operating systems that now exist. As for hardware manufacturers, efficiency of production would be a key determinant of the market success of software com panles. Fuller of DEC warned of a proliferation of standards and called for a more deliberate approach to the issue. "EW]ith too many standards, I believe, we will tie the hands of developers and inhibit innovation," he said. The challenge, as Fuller described it, is to develop a "set of well-selected standards" for hardware, software, and communication areas that are at a mature stage of technological evolution. "The rest of the playing field," areas where ideas and technology are changing rapidly, should be left "open for innovation," according to Fuller. The questions of what to standardize and when to do it are as contentious as the issue of what the specifications for particular standards should be. Are operating systems at a mature stage of evolution, for example, or do significant advances lie ahead? Once in place, would standards make it difficult for new, superior ideas and technologies to gain acceptance and, ultimately, market share? Fuller suggested that effective standardization by many users on UNIX illustrates the mixed consequences of standardization. Tester noted, however, that reducing the cost of equipment is only one factor to be considered in decisions to develop or encourage standards. Noting that the cost-effectiveness of standards will decrease as they age and technology improves, Tester said that companies that"leap ahead" of the standard will be excluded from a major segment of the market the federal government and, perhaps, its contractors. Debates over standardization will continue in domestic and international are- nas. But if the push for standards continues to gain momentum, standardization will greatly influence the business strategies of software companies. Perhaps the only choice for U.S. firms, several speakers suggested, is to join forces and develop U.S. standards that are superior to those now under consideration. "If the Japanese are getting together and making a standard," Reddy asked, "why do we not get together and make a better standard than they do, and always stay one step ahead of the competition?" NOTES 1. U.S. Department of Commerce. 'Computer Equipment and Software," 1990 U"S. Industrial Outlook (Washington, D.C., 1990~; figure provided via personal commu- nication with a Department of Commerce analyst.

OCR for page 30
SOFTWARE 39 2. Dertouzos, Michael L. Richard K. Lester, Robert M. Solow, MIT Commission on Industrial Productivity. Made in America: Regaining the Productive Edge (Cambridge, Mass.: MIT Press, 1989), p. 264. 3. U.S. Department of Commerce. "Computer Equipment and Software," 1989 U5. Industrial Outlook (Washington, D.C., 1989), p. 26-3. Computer Science and Technology Board, National Research Council. Scaling Up: A Research Agenda for Software Engineering, (Washington, D.C.: National Academy Press, 1989~. 5. Computer Science and Technology Board, Scaling Up: A Research Agenda for Software Engineering, 1989. The CSTB has explored intellectual property issues in software through a workshop and forum in the fall of 1989. A report will be issued in mid-1990. National Technical Information Service. "Foreign Technology," Abstract No. 37, 031, September 13, 1988. ESPRIT, the European Strategic Program for Research and Development in Information Technology, is an umbrella program that encompasses some 200 spe- cialized projects. One product that has emerged from ESPRIT is the Portable Common Tool Environment (PCTE), a standardized substrate, or foundation, for developing software for large systems. PCrE accommodates existing and emerging software engineering tools, which permits programmers to exchange tools and researchers to develop new ones that enhance the utility of the software substrate. Eureka, a European Community research and development program focused on commercially promising technologies and innovations, has provided funding for a software factory based on PCTE. 8.