Planning for Personal Computers

Sources, reliable and unreliable

Between 1980 and 1995, networks of personal computers replaced timesharing as the core of academic computing. For much of this period, academic computing and the commercial mainstream continued to follow separate paths, but they eventually converged in the 1990s.

Sources of information about this transformation are scattered and unreliable. Much of what was written at the time described what was planned, which was often not what was achieved. Many of the newspaper articles were university publicity laden with euphoria. The New York Times had a correspondent who was well known for presenting his personal biases in the guise of factual reporting. Perhaps the most reliable source would be the notes of Judith Turner from the Chronicle of Higher Education, but she has wisely kept her notes private.

Much of what has been written subsequently is pure revisionist history. Individuals and organizations describe what they wish had happened. Wikipedia is frequently a good source of technical details, but the contextual information is often tendentious. In practice we went through a decade of hyperbole, resistance to change, blind alleys, failed plans, and spectacular achievements.

Three books published by EDUCOM are a good contemporary source. EDUCOM was the university computing society. Its publications, annual conference, and corporate associates program were an important exchange of information. I was a member of the EDUCOM board for most of the 1980s and its chair for six years. Under two outstanding presidents, Jack McCredie and Ken King, EDUCOM was a major conduit for contacts with corporations and the government.

One day in fall 1980 Jack McCredie and I were walking across the Stanford campus. He mentioned that he was planning a monograph on university computing. Ten universities would each write a chapter about their strategic planning. I casually agreed to be one of the authors without admitting that at Dartmouth we were so busy building our computer system that we did not do any formal planning. This was the first of the three books. My wife edited the other two, which were about campus networking and libraries.

Two of Ken King's creations were the Coalition for Networked Information, which brought librarians and computing center directors together, and a networking task force that lobbied for higher education in the development of national networks. Al Gore is rightly ridiculed for his claim to have invented the Internet, but he was the first high-level politician to recognize its potential and a great supporter of these efforts.

Because I served on numerous committees, I often saw planning papers from other universities. About 1980, Stanford wrote a particularly insightful set of papers on topics such as networking, office automation, libraries, and so on, but, for me, the most illuminating was the Preliminary Report of the Task Force for the Future of Computing at Carnegie Mellon.

Managing the first personal computers

Personal computers posed a dilemma for universities and their computing centers. After years of struggle we had learned how to manage timeshared computers. A few years earlier, some computing directors had resisted the purchase of minicomputers by departments, but eventually they accepted that minicomputers were frequently cost-effective. Now they had to face a wave of incompatible personal computers that were often purchased by people with little knowledge of computing. The total expenditure on computing was going up steadily and it was not clear what the university was gaining. For a generation who believed that every computer cycle was precious, it was painful to see computers sitting on desks to be used for only a few hours every day. Some computing directors thought that their mission was to protect the university from this wasteful proliferation.

Gradually people began to realize that personal computers were more than a fad. Admittedly they could not do many of the tasks that the larger machines did well, but every year they became more powerful and the application programs grew better. For example, in summer 1982, a student and I developed a planning model for hardware purchases using Visicalc on an Apple II. The time and effort were a fraction of what we would have taken on a timeshared computer. Above all people began to appreciate the benefits of having their own machines. I used to compare personal computers to private cars. It might be cheaper and safer for everybody to travel by bus, but individuals pay for the convenience of having their own cars.

The showpiece projects

At the beginning of the 1980s, the academic world was excited by the launching of ambitious computing projects at several universities. The most prominent was the Andrew project at Carnegie Mellon University, which included a massive grant from IBM. Never to be outdone, MIT soon afterwards announced the Athena project, jointly with IBM and Digital. In the same year, 1982, Drexel University announced that every freshman would be expected to have a personal computer, and some time later they put out a press release stating that they had chosen an unnamed computer from Apple (which they described as being supplied with a "house", a typo for "mouse"). This computer was later known as the Macintosh.

Several other universities, notably Brown University and Franklin and Marshall College, had important initiatives, some universities passively accepted the arrival of personal computers on campus, and some universities actively discouraged the proliferation. Where there was no campus-wide project, departments or schools, such as engineering or business, often built their own local networks. Harvard is an example of a university that deliberately held back, waited to see what consensus emerged, and then purchased a very fine network, but Harvard is a rich university and could afford to wait. The pioneers endured a great deal of hassle, but received major grants of equipment and money. Brian Hawkins, a force behind the Drexel program, once observed that the universities that took the lead were usually the universities with strong academic leadership.

Whether students should be required to own personal computers formed a lively topic throughout the decade. At Dartmouth we standardized on the Macintosh in 1984 and at our urging, more than 75 percent of freshmen bought them, but not until years later were students required to have computers. Carnegie Mellon was more typical. Although, the university had a huge project with IBM, no explicit requirements were made to buy computers. The campus store sold personal computers from both IBM and Apple, and workstations from several manufacturers, with deep discounts for hardware and software.

Organization for distributed computing

The story of these large projects is incomplete without a discussion of the struggles that the universities went through to fit them into their organizations. Universities always have great difficulty in coming to a consensus, particularly when resources have to be reallocated, and these projects were no exception.

In a talk in 1985, Steven Lerman, one of the joint heads of the Athena project at MIT, described a problem which he called "the management of expectations." Projects such as Andrew or Athena need great initial enthusiasm to marshal the resources that they require. Then follows a long gestation period before any signs of progress are seen, and the first results are a pale reflection of the dream with which the projects began. Carnegie Mellon is an extraordinarily entrepreneurial university in its willingness to tackle new ideas, but the university is not rich. The growth in computing demanded resources of money and space, which were tough to find, at exactly the same time that the Andrew project was struggling to live up to its expectations. Fortunately, President Cyert was personally committed to the project and the resources were found.

As the importance of computing grew, job titles were inflated. At the Open University in the 1970s, the head of academic computing was simply the manager of student computing. At Dartmouth my title changed from director of computing services to vice provost and at Carnegie Mellon it was vice president. These distinctions seemed important at the time.

Whatever their titles, computing directors were in a difficult situation during the transition to decentralized computing. Some saw their duty as protecting the central computing budget from the inefficiencies of departmental and personal computing. Others allowed the central service to atrophy while they invested in networks and decentralized computing. No university had the resources to do both really well. During the transition almost every university changed its computing director. Sometimes the university forced them out. Others left for better opportunities or because of frustration with the university's administration. I had been drawn to Dartmouth and Carnegie Mellon because of two outstanding presidents, John Kemeny and Richard Cyert. Working for their successors was no longer enjoyable and I moved on.

Until about 1980, most computers were purchased with central funds and a handful of people made all the decisions. Only in rare instances were computers seen as personal equipment. A decade later the position had been reversed. In 1990, the annual expenditures on computing equipment at Carnegie Mellon was about $12 million, with only $2 million coming from central funds. Purchases made with personal funds through the computer store were over $3 million and the rest came from departmental funds or research grants.

People who spend their own money naturally expect to choose their own equipment. Yet, everybody benefits from a certain level of campus-wide coordination. The organizational challenge was to reconcile independence in decision making with sufficient standardization to create a coherent campus environment. At Dartmouth, the campus was dominated by Macintoshes, but there were still large numbers of other computers. At Carnegie Mellon, although no formal commitment was made, the expectation had been that the Andrew project would lead to a largely homogeneous environment dominated by IBM. This did not happen. A tour of the campus in 1990 would have found Apple Macintosh and IBM personal computers everywhere, with hundreds of workstations made by IBM, Sun, Digital, Hewlett Packard, and NeXT. These computers had been selected by many different people at different times with differing needs and resources. Each selected the equipment which best fitted current needs. Yet, at every university, a certain level of standardization emerged because of the difficulty of supporting too wide a range of equipment. The computing environment was so complex that only a few types of computer could be well integrated into it. Most people bought those types of computer.

Gradually universities came to realize that they needed somebody who would oversee the computing strategy for the entire university. This person was not necessarily the person who managed the central computers and did not have to be in charge of both academic and administrative computing. At Carnegie Mellon I had responsibility for academic computing and infrastructure such as networking, but not for administrative computing. This division between academic and administrative worked well. I am surprised that it is not more common.

At Dartmouth, the Kiewit Network and the Macintosh project were creations of the computer center and there was never any discussion of creating a new organization. Andrew at Carnegie Mellon and Athena at MIT both began as separate organizations outside the existing computer centers, but thereafter followed different paths. Athena remains a separate unit, but at Carnegie Mellon I transferred the responsibility for deploying Andrew to the old computer center. This required a painful reorganization, but the organizational framework has survived.

The institutional structures for academic computing that have emerged at most universities usually parallel the organization of the rest of the university. The organization that I see today at Cornell is typical of large research universities, partly centralized and partly decentralized. It combines aspects of both extremes: entrepreneurial faculty with powerful and independent deans, and yet a fairly strong computing center that provides shared services such as networking and email to the whole university. This central organization is also responsible for the university's administrative data processing