from Jack Palmer March 12, 2009
I have seen some recollections of the subject in recent correspondence.
Software product testing (product testing later came to be called system assurance) began in mid-1961. IBM's first programming conference was held in New Hampshire in June of that year (later called "the Bald Peak meeting" after the venue: the Bald Peak Colony Club in Melvin Village). On the last day the product division presidents issued a number of directives, including that their product testing organizations would begin to test programs.
In Endicott working on 1401 Fortran later in the year, I remember meeting with Dick Carlson of the laboratory product testing department, who had been assigned (perhaps with others) to test that program. I think George Weisert was the manager involved; Hugh Harvie may have been the second level manager. By the time of the System/360 alpha tests in early 1964, Tom D. Hoffman was the manager.
There is more information on the Bald Peak meeting in C. J. Bashe et al., 1986: IBM's Early Computers (Cambridge, Mass.: The MIT Press).
A highly recommended book including developement of the 1401 is IBM's Early Computers by C. J. Bashe et al., 1986. Ed Thelen got so excited that he bought three of them. Sample text
from Gary Mokotoff March 13, 2009
I don’t know if you consider it a legend or a story but the original Autocoder developed by John Wertheim and myself in 1960 was “product tested” by giving what we would call today a beta-test version of Autocoder to the IBM Datacenter in the basement of the Time/Life building and they would periodically tell us of bugs they found.
Sample text from IBM's Early Computers by C. J. Bashe et al., 1986.
368 Chapter 9
9.10 A Conference on Programming
Fundamental to the concept of the general-purpose digital computer was its adaptability to a wide variety of computing tasks. The instrument of that flexibility was the program, whereby the capabilities of the machine were marshaled and orchestrated to accomplish a specific purpose: the processing of a specific set of data to produce a new and more useful set. A data processing system useful to any enterprise necessarily included both machines and programs. During the 1950s IBM came to realize through its own undertakings and those of its customers that, for any general-purpose machine, programs could be written that were themselves general purpose in nature. The most useful were of two major types: those designed for processing-oriented tasks, such as sort and utility programs, and "system programs" designed to make programming and operation more efficient, such as assemblers, compilers, and supervisory monitors.
General-purpose programs, being amenable to exchange among users and to cooperative development, provided a basis for the formation of SHARE and GUIDE. But such programs were also plausible as components of data processing systems offered by vendors, for the same reason that the machines themselves were: the savings achieved through their use were substantial. The availability and quality of such programs had come to be one of the factors by which users differentiated among competing machines. This was very apparent to IBM's programmers, since they had been involved in most of the technical contacts with users throughout the 1950s. By contrast, the programmers knew little about IBM's product-development procedures until the latter portion of the period, when the increasing complexity and variety of machine components and the need for product information earlier in the development cycle forced them into frequent contact with their engineering counterparts. As a result, they began to perceive that program development lacked many of the disciplines and supporting services that aided development engineering. The uncertain availability and distant locations of machines on which to check their programs was one example. The exclusion or token inclusion of programs in IBM's product division plans was another. Job codes and position descriptions were still provisional. Programmers felt, in short, that IBM's approach to programming was inconsistent with the significance of programs.
In early 1961 these perceptions were of concern to David Sayre, the first incumbent of a new position on the corporate staff director of programming. Sayre had participated in the development of the
704 FORTRAN compiler and had written the October 1956 FORTRAN manual. His manager, who a few months earlier had acted to meet related needs in the sales division, had suggested that Sayre organize a conference of IBM programmers with the purpose of synthesizing their views on what should be done. Sayre enlisted a FORTRAN colleague, Ziller, to help. They organized and issued invitations to a conference that would divide into workshops, for which chairmen were selected, to develop objectives and recommendations for achieving them.
Learson, meantime, had received from a customer an unsparing criticism of the quality and timeliness of IBM programs. Learning of Sayre's plan, he made a forceful case that something should change as a result, insisting that workshop chairmen be named from the ranks of product and marketing division executives. He further directed that presidents of the relevant IBM divisions appear at the conference for the last day or two, receive recommendations, and make firm action plans before leaving. The mid June conference, held at the Bald Peak Colony Club in Melvin Village, New Hampshire, began with a speech by chairman of the board Tom Watson, Jr., that acknowledged the increasing importance of programming and noted the expensiveness of scheduling failures. Expressing confidence that the assembled group could prescribe means to fix the problems, he urged it to do so.
The temperature rose to over 90 degrees in New Hampshire early in the week and stayed there; windows and doors were opened wide and informal attire was in order. The workshops met their schedules to set down by Tuesday evening objectives, recommendations, and supporting data. At plenary sessions on Wednesday, the chairmen presented their results. On Thursday and Friday the presidents of the sales division and the two product divisions studied the results and drafted an "Action Program," a tersely worded document setting out some thirty-five actions. The actions included a corporate-wide career ladder and recognition and reward system for programmers, recognition of programs in divisional plans, expansion of the product-testing departments' responsibility to include programs, and increasing by over 50 percent the corps of sales division programming specialists. Also included was the following action: "The product divisions will establish unfettered `programming technology' organizations to pursue advance programming concepts, applied research, automatic program production techniques, etc. by August 1, 1961. The Bald Peak conference, IBM's first major programming conference, led to ...