Bob Bemer returns from Paris to Phoenix --
Still with General Electric

Checking In at Phoenix

My employee number, when reassigned from Paris to Phoenix in May of 1966, was "SP000". One guesses that the "SP" stood for "SPecial", and indeed I was, still more or less with a direct line to Dr. Louis Rader. Later, when I was regularized, it was "00019", the lowest there. One of the later General Managers, Richard R. Douglas of Honeywell, felt irked because it was lower than his, such that his name always appeared second on the payroll summaries.

Note his initials. Later, when Sen. Barry Goldwater was to appear at a dedication for a new software building, I contrived to find a man who fielded a robot that looked like R2D2. So Dick Douglas billed himself as R2D1. But to go back --

At first I was assigned to work for Dr. John Weil, who was a bit suspicious of me then. His managers always thought I was going to be the manager for all GE software when someone got around to fixing the organization. This never happened. To the very end (when Honeywell purchased the operation in 4 more years) there were something like 10 autonomous software groups (including Italy and France), and the only attempt at overall control was an admonition to work together. GE had one and only one dimension of management.

John Weil was actually a very fine man, with a topnotch brain and traits of an educated gentleman. Plus outstanding integrity. But our first tangle had come when I was still at Bull GE. One of his projects was a new tape operating system for the model 400. This when a disk unit was generally available. I have no record of who talked him into it.

I didn't believe this could be a viable product, so I met Charlie Lecht of ACT and his lieutenant Ralph Stout in London to see if they would be interested. By a coincidence I was reading a British newspaper while going there, and one of the comic strips was Buck Rogerish. One panel had the hero saying "Let me take you to the disk" (meaning a flying saucer-like vehicle). This was too much of an omen. I showed it to Lecht and Stout, and gave them the contract.

Weil wanted to cancel that Disk OS for the Model 400, but I argued with Dr. Rader to keep it going because 1) the inhouse Tape OS was in poor shape, and 2) everyone now knows that tape-based could never compete with disk-based systems.

Someone to Finally Jump

Wengert called me in after many other attempts had fizzled (just ignored, I think), and asked me to go there to see what I could do for amelioration. The problem was that there was a big nationwide airline strike on just then. So I went to the Phoenix airport on a Friday morning in July of 1967 and waited until I could find a flight going East. Wherever that was, I did the same again at that airport.

My itinerary turned out to be: Phoenix - El Paso - Dallas - Memphis - Nashville - New York - Albany, followed by a rental car to Schenectady. I could not get to the plant until the mid-Friday-afternoon. But I worked fast, found what most of their problems were, and left late the same day (a visit of no more than 3 hours). The return of Albany to Phoenix by way of New York, Dallas and Los Angeles is not your most direct flight, either.

When I saw Wengert on Monday, he said to me after my report: "You're the first person around here that, when I asked them to jump, actually jumped!" I thought then, and do now, that this avoidance of obedience was rather generic in GE, because when managers are always kept moving around to different places and assignments they get to know a lot of buddies to depend upon for support. I'm sure it was contributory to GE's resigning the computer business.

The New Science of Software Measurement

Dr. Weil, as a manager, had more faith than I did in the principle of "tell them what you want done and wait for it to happen correctly". So he kept getting sharp notes from higher management about certain software that was not meeting specifications -- particularly the COBOL processor for the 600 line. That was it. He had been wondering what to have me do, and here was an interim assignment that would take "my" measure, too.

I found a couple of men, Leroy Ellison and Harry Cantrell, not being utilized to their total ability. They liked the scope and freedom I proposed for the new project. We started on COBOL measurements, finding some remarkable problems. One was of 1200 patches that never got around to seeing a reassembly. Nobody had ruled that periodic reassemblies should be made. See that separate story.

Ellison and Cantrell were so successful that I pointed them at the GCOS operating system, too. This was a very early software measurement effort. I don't remember that anyone else, other than Ken Kolence of the new Boole and Babbage firm, was doing it. The results were remarkable. Many had theories about where time would be wasted in an operating system, and hardly any of these were near correct. I thought the results very significant, and arranged for Leroy to give a paper at the next IFIP Congress in Edinburgh in August of 1968 [1].

Some may remember Ellison and Cantrell for their work after they left GE, when Ellison founded CAPEX. Whence the first optimizer for COBOL- generated object programs. And that turned into ViaSoft.

Irresponsible or Incompetent Market Planning

I remember sitting several times in marketing meetings where engineering had proposed a new product. Statistics would be brought out about how many such could be sold in the next 6 months. The consensus was often "Well, that justifies it -- let's go ahead".

I had a problem, according to these people. I would ask, after the 6 months had gone by, "How many did you actually sell?" I don't know what it was about these folk. They thought I was rude to even ask such a question, especially when the answer was often "none"! This was very strange for me, especially having once worked for IBM, where people with such bad predictive capabilities would not survive in their jobs!

Software Production Standards and Requirements

After the Ellison/Cantrell successes with measurement, I started to think about the larger aspects of environments and rules for total software production. I had made a start on this at UNIVAC, where I installed, with the help of Milt Bryce, the first labor distribution system ever applied to software production.

I made up categories or stages of such production, and postulated a series of questions in each. With a ranking or scoring for each result. I compared General Electric's software production with what I knew of IBM's and UNIVAC's (a considerable portion of the world in those days).

When I finished I sent the results to Dr. Rader, because GE did not fare at all well in the comparison. I added that he needed to come to Phoenix more often, and see for himself. I suggested special monitors from the outside, much as I was. To no avail. The picture I painted was so bad that he was sure it could not be so. In any case, he ignored some pretty dire warnings. Perhaps someone else was telling him false good news.

So one day everything broke loose. Hilliard Paige relieved Rader of his overall jurisdiction for computers within GE. It was a bad day. I shall never forget that night, with Rader and Bob Curry sitting around our pool, with my stepson, who was then assistant to the President of Yale University.

I was resolved not to waste my efforts. In 1968 John Weil received an invitation to attend the First NATO Conference on Software Engineering, at Garmisch-Partenkirchen in Germany, an event of seminal importance in the software profession. For some reason John Weil was unable to accept, so I was proffered as a substitute.

Thinking my recent work most appropriate to the topic, I took the documents along, and made them widely available. You can find them printed in near totality in the Report of that conference. And thus was the concept of the "software factory" born. [2 through 9].

Important Contributions

Despite management, I would insist, GE made some great technical developments. Among these were:

• The development of IDS, the Integrated Data Store, by Charlie Bachman, aided by Susan Brewer. GE did not really capitalize on it, and I said so vehemently when protesting the proposed abandonment of the GE 600 architecture to copy that of IBM. (I wanted to jack up the software and put a new hardware engine in the body, a la the 1107 to 1108 change I championed at UNIVAC). I don't think many people understood Charlie's vision of the future when databases would be more important than the programs that manipulated them.

• GECOS, the 600 operating system (when finally tuned), which had features far superior to competitor IBM. Among other values, when it became operational on the GE 6000, General Electric's operation in Schenectady, under Paul Naegele, found that it took only 9 people to run the computer under that operating system, as opposed to 30 needed to run an IBM computer of comparable computing capability. Significant? You bet! Especially as hardware costs went down and people costs climbed. I wanted Honeywell (at that time) to take out a big ad in the Wall Street Journal to stress this significance. They didn't get it either.
• The total Multics development, in cooperation with MIT, was a waste, eventually. Here they were, miles ahead of IBM with a timesharing system and a fine cooperative development team. But they never found out how to heal the chasm between that and their standard products, despite my proposals to seed each side with programmers from the other.
• The line editor associated with Multics, and taken over for the 600 line as RAES (Remote Access Editing System). I once invited a group of distinguished visitors from the UK to join me at my house for libations. While they were there I demonstrated the remote editor from the Model 33 Teletype I had at my home. They were greatly impressed. Especially Dr. Stanley Gill, one of the three authors of the very first book on the art of programming. (You can mark this down -- in 1966 or 1967 -- as making me one of the very first to use a home office for computers).

In preparation for a periodic reunion of former GE Computer Department people, someone sent out a questionnaire that included the following (with my replies):

Question: What do you feel was your greatest contribution to the GE Computer Department?

• Although dropped by management, even before the sale to Honeywell, the work I did to define the software factory. Much of the story was presented at the First NATO Software Engineering Conference in 1968.
• The ACM 70 program. Although they did not know it was a contribution to the department, it was, in terms of publicity and market focus.

• They assigned too few managers that understood the business. E.g., Erv Koeritz, one of the PHX GM's, on COBOL problems -- "I know the chemical business - I can get you plenty of cobalt".
• Over the entire time I knew them, some of the non-managers were outstanding. Middle management was cliquey and largely incompetent. Top management seldom understood the computer business.
• The product type was not suitable to their accounting system. Too long a product cycle, too invisible a product.
• The COPY-IBM effort was just the last straw. It thrust the computer business into critical analysis, and the dollar figures (even as much as they were understated) were just too overwhelming for GE to go that route without guarantees and reassurances.

• In some senses. They pioneered in communications, timesharing, and banking. Charlie Bachman led them to glory, but they never capitalized on it in databases.
• Advertising was poor, though. And they never really differentiated between good and bad assets. They did not recover from Haanstra's death. UNIVAC-like, they snatched defeat from the jaws of victory. Innovations were left to wither and make profits for others.

1. R.W.Bemer, A.L.Ellison, "Software instrumentation system for optimum performance",
   Proc. IFIP Cong. 68, Stwe 2, Book C, 39-42
   -- Computer Abstracts 70-1573
   -- Computing Reviews 15765
   
2. R.W.Bemer, "Software systems customized by computer",
   Proc. IFIP Congress 1965, Vol. II, 356, 1965 May 24-29
   
3. R.W.Bemer, "Economics of programming production", in
   Economics of Automatic Data Processing, A.B.Frielink, Ed.,
   North Holland Publ. Co., Amsterdam, 1965, 155-166
   -- Computing Reviews 09742
   
4. R.W.Bemer, "Aspects Economiques de la Production de Software", in
   Mecanographie et Informatique, 1966 May
   
5. R.W.Bemer, "Economics of programming production",
   Datamation 12, No. 9, 32-39, 1966 Sep
   
6. R.W.Bemer, "The economics of program production",
   Proc. IFIP Congress 68, Booklet I, 13-14
   -- Computer Abstracts 70-1409
   
7. R.W.Bemer, "Machine-controlled production environment",
   Report NATO Conf, on Stwe. Engg., Garmisch, 94-95, 1968 Oct 7-11
   
8. R.W.Bemer, "Checklist for planning software system production",
   Report NATO Conf. on Stwe. Engg., Garmisch, 165-180, 1968 Oct 7-11
   
9. R.W.Bemer, "Manageable software engineering", in
   Software Engineering 1, Proc. COINS III, Academic Press,
   New York, London, 1970, 121-138
   -- Computing Reviews 21123