For background on a story, reporters at The New York Times went to the morgue, the paper’s file of clippings. Around 1969, IBM’s Federal Systems Division began building the Times something new to replace it: an online system called the Information Bank. Reporters and editors at 40 display terminals could pick index terms, scroll through abstracts and call up the full article from a microfiche device fitted with TV cameras. The Times planned to sell access to outsiders, with room for up to 500 remote terminals.
IBM organized the work in an unusual way. Instead of dividing the system among many programmers, it built what it called a chief programmer team, “similar in concept to a surgical team.” Members were specialists who assisted the chief programmer, the project’s lead, IBM’s Terry Baker wrote, “much as nurses, anesthesiologists and laboratory personnel assist a surgeon.” The chief was “both the prime architect and the key coder of the system.” A backup programmer worked alongside, ready to take over if the chief left. A librarian kept every program and test result in order. An analyst helped, and a few more programmers joined later to write the rest “under the close supervision of the Chief and Backup Programmers.”
The idea came from IBM’s Harlan Mills, who had earlier written a large system largely on his own. At a 1969 NATO conference, an IBM manager, Joel Aron, called Mills “a very unusual individual,” with few equals among IBM’s 2,000 programmers. Aron also said the newspaper project was going “substantially slower,” partly because users “don’t really know what they want.”
The Times prepared the acceptance tests itself. The file maintenance part, 12,029 lines of code, was tested against 171 separate functions over a week. “No errors at all were detected.” Across the whole system, 83,324 lines, testing found 21 errors. Once the system was running, another 25 had turned up by the time of Baker’s report, which Baker put at about three errors per 10,000 lines of code. The programs written by the chief and the backup, he wrote, “had about one error per year of effort.” The whole project took 132 man-months.
Baker was careful about what it proved. “The team, in this experiment, was a relatively experienced one,” he wrote, “and it performed at an above-average level.” He believed such teams “should probably be able to double normal productivity,” but the comparisons were informal. Years later Mills looked back on what happened when other companies copied the idea with the wrong people as chiefs: such teams “did not always work out well, and some led to real disaster.”