01/02/03
Pages 242-256
In 1974, a task force recommended that Xerox form a new division, the Systems Development Division (SDD), to make advanced office "systems" for sale to large corporate customers, a market then dominated by IBM. The advanced office system, ultimately called the Xerox Star, would have personal workstations with bitmapped display screens for every individual, as well as servers (communal machines) for printing and file storage. All were to be linked to each other by Ethernet and run PARC's most advanced software. The proposed system was to be a scaled-up, commercial, version of the Alto.
The target users of the advanced office systems were not secretaries and clerks but their bosses, who were executives and professionals. Marketing experts of the time were skeptical about this decision because, at that time, secretaries and clerks performed the mechanics of almost all office work. The market concept of aiming at professionals proved consistent with the office revolution that eventually occurred, after which professionals do much of the mechanical work on documents and information with their own computers. That change did not occur as soon as Xerox had hoped, however.
Bob Metcalf, who earlier developed the Ethernet with David Boggs, had been lured away from Xerox PARC with the promise of more money (20-30%) and a Vice Presidency within 6 months at Citicorp, which handled electronic fund transfers for Citibank. When Citicorp balked at fulfilling the promise of a vice presidency because Metcalf was not yet 30 years old, he was ready to leave. So, Xerox attracted him back to work in the new SDD, but with a much higher salary (10-15% increase). The strategy of extracting a raise from Xerox by working somewhere else for a time became known as a "Metcalf promotion."
Metcalf had two main responsibilities:
Increase the speed of the internet from 3 to 10 megabits per second (our wireless LAN is 10 Mb/s and the wired LAN is 100 Mb/s), and
Oversee the design of the Star's central processing unit (CPU).
As far as the CPU was concerned, Metcalf had two problems:
The personal relationship between Metcalf and Chuck Thacker, who was designing the CPU, was awful, dating back to the first time Metcalf had worked at PARC, and
The CPU design was in trouble.
Thacker had basically scaled up the design of the Alto CPU, but the CPU for the advanced office systems had to handle not only word processing and networking, but also high-quality digital copying and printing with high-speed laser printers. Thacker hit a brick wall: the design was not really good for either purpose - it was too big, too slow and too expensive. He went back to CSL, his home group at PARC. Even as outstanding an engineering as Thacker does not always succeed.
Butler Lampson came up with the key idea of feeding a fixed number of instructions through the CPU during a single clock cycle. The increased predictability of the CPU's work schedule of this "synchronized" approach led to smoother communication between the CPU and the other components such as the Ethernet transceiver and the disk controller. Consequently, the size of the memory buffers needed to hold data from the peripherals while waiting for the CPU to need it decreased dramatically, which consequently helped solve the size and cost problems.
The various difficulties delayed introduction of the Xerox 8010 office system, the "Star," as a commercial product until April 1981. It was a great technical achievement, but as we'll see later, its success as a product proved disappointing.
SDS, the computer company that was supposed to bring Xerox into the computer age when Xerox bought it in 1969, had been folded into Xerox as a division in 1972 to hide its losses. In 1975, the Xerox Corporation itself was not doing well, and Xerox concluded that the SDS division could neither be saved nor sold at any price. To cover the expense of getting out of the business, Xerox took an $84.4 million write off, and posted its first annual loss in 15 years.