Whatever Happened to the Seven Dwarfs Dwarf Five: General Electric

by John C. Dvorak

It was a coinage of the mid-1960’s as IBM dominated the computer business. IBM and the Seven Dwarfs was how the business was described. By 1965 IBM had a 65.3-percent market share of the industry. The seven dwarfs shared the rest. They were: Burroughs, Sperry Rand (formerly Remington Rand), Control Data, Honeywell, General Electric, RCA and NCR. The following is the story of the so-called King of the Dwarfs: General Electric (3.7-percent share in 1965).

If you go to the General Electric website and attempt to find the history of the company’s foray into computers you simply won’t find it. Weirder still is that the entire episode is completely missing from the company’s timeline at www.ge.com/ibhist2.htm.

It’s as if GE never made a computer, let alone a series of mainframes. It’s somehow lost, forgotten or repressed. The company was called the King of the Dwarfs not because it actually competed well in the computer business, but because it could have been a contender — if it had the will. It didn’t.

The irony to the GE computer era was that, due to circumstance more than anything else, the company is, oddly enough, as responsible for the current computer revolution as is any company we like to credit. One can easily argue that todays modern operating system architecture traces directly back to the GE mainframes and the development of Multics, the first modern computer operating system, still in use today.

GE began life in 1878 as the Edison Electric Light Company. The original purpose was to support Thomas Edisons lightbulb research, his having already invented the stock ticker. A year later he got the patent for an incandescent bulb and in the process invented the Edison dynamo. It was soon on to electric streetcars and an invention center. The company soon became Edison General Electric which merged in 1892 with Thomson-Houston to become General Electric. The phonograph and motion pictures were among the Edison inventions. Even after the Edison reign was over (he died in 1931) the company kept inventing things. The electric locomotive, steam turbines and the first US designed jet engine. There was no connection with computing as the field evolved except that GE made many of the tubes used in the early computers.

It was 1956 when the Bank of America requested bids from the computer industry to design and build a new computer system for its operations incorporating new ideas for banking. GE, almost out of the blue, bid on the job. This wasn’t an easy decision since the president of GE, Ralph Cordiner was at odds with Doe Baker, the VP of the electronics division. Baker wanted to get into the mainframe business knowing that with GE’s massive resources it could be the king of the industry. Cordiner wanted no part of it. In fact it’s believed that the only reason Cordiner let Baker’s group make a bid was because Cordiner was assured by his staff that IBM would win the contract. But GE won and benefited with a $31 million contract — then the largest ever received for a non-government computer order. The same contract eventually paid out $50 million. While this deal was profitable, Cordiner was against any other computer deals. For the next four years there was a lot of political infighting as Cordiner finally relented in 1960 and let the company go into the computer business with products other than the custom made system, the ERMA (Electronic Recording Method of Accounting) computer system made for the Bank of America. This is the system, incidentally, which introduced the use of MICR (magnetic ink character recognition) — those little characters we now see on the bottom of ALL checks.

In the early 1960’s as the GE computer business was starting to boom a new thought crept into the arena: multi-tasking and remote access. MIT and Dartmouth were both addressing the problem and developing ways of letting the computer do something other than batch processing. This thinking led to the formulation of a be-all/end-all killer operating system called Multics (Multiplexed Information and Computing Services). This effort as well as MIT’s project MAC (Multiple Access Computers) brought ARPA money into the picture and can be seen as the origin of the Internet with the government helping to fund this early research in remote access and time-sharing.

Multics needed a computer on which to run and GE with its 635 (modeled after the IBM 7090) seemed to be the ticket. It evolved into the 645 for Multics. This was a 36-bit word machine with an execution speed of about 435 KIPS. It would later employ a then massive 4 megabyte magnetic drum for storage. With it GE was now at the center of the revolution in computing. IBM missed this boat because the architecture of Multics, which included numerous concepts which we take for granted today, did not interest IBM. Some of the things that came out of Multics, for example, were dynamic linking, pages and segmentation, shell models and the real anathema to IBM at the time: the use of upper and lower case. Even todays modern file system derives from Multics.

Multics was first proposed in 1963 and continued development during the decade. One of the early sponsors Bell Labs dropped out around 1969 and took what it learned to develop UNIX, a relative subset of Multics. Just as Multics was ready to take over the world riding on a GE mainframe, the batch processing side of the business revolted still preferring to sell GE’s batch process OS dubbed GECOS (General Electrics Comprehensive Operating System). MIT, meanwhile, began to sell full-time use of the revolutionary new Multics operating system in 1969. Everyone was crazy for it. During the decade GE’s computer division became the GE Information Systems Group. It had 25,000 employees and $1.5 billion in installations. By the end of 1969 the company told Wall Street, “In terms of overall business performance, we are in the best shape of our history.”

A few months later, just as things were heating up with Multics, GE sold the entire computer business to Honeywell. It left the business as quickly as it entered.

Multics was the first OS to promote remote terminal access for all computing, have the computer run continuously, use an internal file system, structure the management of the machine, accommodate large and small users equally well, utilize dynamic linking and manage multi-level memory — in short, do what computers should do. Some observers felt that GE simply couldn’t see a computer doing much more than banking functions: add and subtract and sort checks. Once confronted with other possibilities they simply had to get out of the business. It was something GE couldn’t handle. Others say that it was a bottom line issue. As the new style of computing came along it looked too expensive a proposition. Multics took 6 hard years to develop and it was still buggy. In hindsight and with the eventual decline of the mainframe business, GE did the right thing although old-timers lament the decision preferring to believe that IBM could have been beaten and decimated right then and there. Wishful thinking methinks.

A final note regarding the effect of Multics on the world in which we work today. The seminal program of the 1970’s and one of the most important software packages ever developed, VisiCalc, was developed on a Multics machine by Bob Frankston.

Links:
Other Dwarfs: Burroughs, Sperry Rand (formerly Remington Rand), Control Data, Honeywell, General Electric, RCA, NCR