Data General

» see dg.com as it lived, on the wayback machine

~ the last words ~

In 1984, the year Data General passed a billion dollars in annual sales, I was one year old. That is the honest distance here. The company was already finished as a force before I had touched a keyboard, and by the time I had a modem it was a name in older books rather than a thing anyone I knew bought. What survived into my working life is the story of one project, which is a strange kind of afterlife for a company that shipped hardware for thirty-one years.

It started as a walkout. Edson de Castro was an engineer at Digital Equipment, where he had worked on the PDP-8, and he had a design for a successor that Digital's management would not build. He left in 1968 with Henry Burkhardt III and Richard Sogge, picked up Herbert Richman from Fairchild for the sales side, and set up in Hudson, Massachusetts. The Nova arrived in 1969: a 16-bit minicomputer starting under four thousand dollars, small enough and cheap enough to end up in laboratories and schools that could not have justified a Digital machine. It sold for a decade.

The seventies belonged to them. Data General was the loud competitor, the one that ran an advertisement about IBM entering the minicomputer market with the line "the bastards say welcome" underneath a headline, an ad that reportedly never ran but hung framed in de Castro's office anyway. The Eclipse followed the Nova in 1974. By 1984 the company had a billion dollars of sales and something like seventeen thousand people. Then the ground moved. Workstations from Sun and Apollo did technical computing better for less money, and underneath them the IBM-compatible desktop took the low end away. The minicomputer as a product shape stopped making sense.

The last decade was a managed retreat. AViiON in 1989 moved the company onto Unix and RISC, and CLARiiON storage arrays turned out to be the only late product anybody wanted badly. EMC announced a $1.1 billion purchase on 9 August 1999 and completed it on 12 October. Every product line was shut except CLARiiON, which EMC sold under its own name until 2012. The company that started because Digital would not build one man's machine ended as a storage acquisition.

~ the record ~

~ the book ~

In 1980 Data General needed a 32-bit answer to Digital's VAX and did not have the budget to do it properly, so a team under Tom West built one on the side in about a year. Tracy Kidder was in the building for it. The Soul of a New Machine came out in 1981, won the Pulitzer and the National Book Award, and became the book that engineers hand each other. The machine was the MV/8000. The project name was Eagle.

That book is why the name is not gone. Almost nobody under fifty has used a Data General computer, and a good number of people who write software for a living can still tell you what "signing up" meant on the Eagle team and why the debugging chapter is the good one. It is a rare case of the documentation outliving every product it documented.

~ what actually went wrong ~

Not the engineering. The MV/8000 shipped and worked, AViiON was a competent Unix line, and CLARiiON was good enough that a much larger company kept selling it for thirteen years after buying the corpse. What went wrong was that Data General was built for a market defined by the gap between a mainframe and nothing, and by the mid eighties that gap had two cheaper occupants in it.

Digital died of the same thing, more slowly and from a greater height. The minicomputer companies of Massachusetts were all optimized for selling a room-sized purchase to an institution, and the customer they had trained stopped existing when a department could buy its own machines out of a departmental budget. Data General saw it and moved to Unix boxes anyway, which was correct and about four years late.

~ the next chapter ~

EMC got a storage business and a customer list. The Nova's instruction set lived on for a while in embedded work and in the Nova-derived designs that a handful of clone builders shipped in the seventies. The Hudson plant went into EMC's hands, and the name came off the buildings.

For anyone who came up on PCs, the useful thing Data General left is a shape you can recognize later: a company that is right about the technology, executes well, and is standing in a market that dissolves under it. That happens more often than outright failure and it is much harder to see from inside. The Eagle team worked murderous hours to build the right machine for a category that had maybe six years left.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »