Oral-History:Jean Sammet

About Jean Sammet

Jean Sammet was a computer pioneer who worked on mathematical analysis problems using an analog computer at Sperry, and in 1955 she became one of the programmers on Sperry's digital computer, the Sperry Electronic Digital Automatic Computer (SPEEDAC). Later in 1955, Sperry and Remington Rand merged into Sperry Rand, and Sammet began to work with Grace Hopper and the UNIVAC.

Sammet left Sperry in 1958 to focus more on computers, and joined Sylvania Electric to lead software development on Sylvania's MOBIle DIgital Computer (MOBIDIC). In 1959, Sammet, along with Howard Bromberg, Howard Discount, Vernon Reeves, William Selden, and Gertrude Tierney, designed the influential COBOL language, as an attempt to create a portable data processing programming language.

In 1961, Sammet joined IBM where she developed FORmula MAnipulation Compiler (FORMAC), a FORTRAN-based computer mathematics language, and researched modelling programming and mathematical languages on the use of restricted English notation.

As an active member, and later Fellow, in the Association for Computing Machinery (ACM), she became the organization's first female president, serving from 1974 to 1976. In 1965, she founded the ACM's Special Interest Committee on Symbolic and Algebraic Manipulation (SICSAM). She was the recipient of the ACM's Lovelace Award in 1989, the 1997 SIGPLAN Distinguished Service Award, the 2009 IEEE Computer Society Computer Pioneer Award and the 2013 NCWIT Pioneer Award. Sammet was awarded an honorary doctorate from Mount Holyoke College in 1978.

In this interview, Sammet discusses her work at Sperry, Sylvania, IBM, her publications, and her role in ACM.

About the Interview

JEAN SAMMET: An Interview Conducted by Janet Abbate, IEEE History Center, 29 August 2001

Interview #625 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

Jean Sammet, an oral history conducted in 2001 by Janet Abbate, IEEE History Center, Piscataway, NJ, USA.

Interview

Interviewee: Jean Sammet

Interviewer: Janet Abbate

Location: Jean Sammet’s home in Chevy Chase, MD

Date: August 29, 2001

Early life

Abbate:

This is an interview with Jean Sammet on August 29th, 2001.

I always begin by asking when you were born and where you grew up.

Sammet:

I grew up in New York City, in the heart of Manhattan.

Abbate:

And you were born March 23rd, 1928.

Sammet:

Yes.

Abbate:

What did your parents do for a living?

Sammet:

Both of my parents were lawyers. My mother stopped working as a lawyer when she got married, but she went to law school at a time when very few women went to law school.

Abbate:

Did you have brothers and sisters?

Sammet:

I had a younger sister.

Abbate:

What did she end up doing?

Sammet:

She was in social work—for the City of New York, primarily.

Abbate:

When you were kids, did your parents encourage you to have a career and get an education? Was that expected?

Sammet:

A career was—I don’t know what they expected; I just took it for granted. This may forestall another of your questions, but from the time I was seven years old, I wanted to be a math teacher, even though I didn’t even know the word “mathematics.” I wanted to teach, and I liked that number stuff! Although my parents were lawyers, and many people told me I should be a lawyer, I had zero interest in doing that; because the work my father did had to do with wills, and estate, and probate, and I can’t think of anything more boring. So I was interested in mathematics from the time I was seven years old, even though I didn’t know the word.

Abbate:

So you were just playing around with numbers, even at that age?

Sammet:

I liked that arithmetic stuff. When I got to high school, I took every math course that the high school offered; and I went to college as a freshman knowing I was going to major in math, which I did.

Abbate:

What kind of secondary schools did you go to?

Sammet:

I went to Julia Richmond High School in New York, which was an all-girls’ school. It turned that out at that time, I really didn’t have any choice as to what high school I went to. There was one other one that was much further away, and I think was also a girls’ high school. Unfortunately, the Bronx High School of Science, which is obviously the place I should have gone, didn’t take girls at that time. It turns out very strangely that at the time, the one alternative was Hunter College High School, which was also for girls, which was competitive, and I’m sure I could have gotten in—but I told my parents I didn’t want to go there, because I didn’t want to have to work so hard in high school. And I remember, my senior year in high school, being up at two o’clock in the morning to finish a history paper, and thinking, “Well, it couldn’t have been any worse at Hunter!”

But the interesting thing about Julia Richmond was that they had a school within a school. That is to say, there was a very large population of girls, most of whom were not academically inclined—they were there to get whatever kind of high school degree they could get—but there was within it a sub-unit called the Country School (don’t ask me why, because I don’t know) that had, for each grade, what I would characterize as the smart girls and the ones that were college-oriented. I think I would have had a dreadful time, had I been in the general population, but this was a sub-school that had excellent teachers. Some of the others were excellent as well, but at least ours knew that the girls there were college-bound, and so we did all the right things. I don’t know whether you’re aware of it, but at that time—and this is of course a long time ago, sixty years ago—New York State had what they’re now trying to do, namely a set of standardized exams for almost every significant course. There was only one English exam. They were called Regents.’ It’s interesting, because all of a sudden everybody’s discovering standardized testing. I went through all the Regents’ Exams, of course.

Mount Holyoke and University of Illinois

Abbate:

How did you end up at Mount Holyoke?

Sammet:

I decided I wanted to go to a girls’ school, and I looked at all the catalogs from all the girls’ schools, and I decided Smith and Mount Holyoke seemed to have the best math courses—because I was prepared to be a math major. I mean, I didn’t understand half the things they were listing, but somehow those two looked more impressive. I also applied to Cornell, just sort of as a safety factor. It was based primarily on the catalogs, because when I was applying, which was 1944, this was in the middle of World War II, and people did not go around making pilgrimages to potential colleges; and I did not know anybody who had gone to Mount Holyoke or Smith. So I just applied.

I must admit, one of the more amusing stories that my mother liked to tell was that in April, or whenever it was that the notices came out, the mail came in just before I left for high school, and I came up one morning beaming, because I had the letter and I had been accepted at one of them; I don’t remember which one. Two days later I came up with another letter in my hand, crying, and my mother said, naturally, “Gee, I’m sorry you didn’t get in!” I said, “But I did!” And she said, “Well, why are you crying?” I said, “Now I don’t know which one to go to!” [laughs.] I picked Holyoke. I’m not even sure why I picked Holyoke, but I am very thankful that I did. It was a great college—still is.

Abbate:

So you went off and got a B.A. in mathematics.

Sammet:

Yes. And took all of the courses needed to get a license for teaching high school math in New York. So I had not only the math major, but the education courses needed for a license. But I minored in political science, for no reason except that I liked it!

Abbate:

Now, when you graduated from Mount Holyoke, were you then qualified to teach math?

Sammet:

Yes. But I went to graduate school right from college.

Abbate:

Why was that?

Sammet:

Why did I go to graduate school? I guess by then, even though I still wanted to teach high school, I . . . To be very honest, I don’t remember. I guess I just thought that was the next thing to do.

Abbate:

And you ended up at the University of Illinois.

Sammet:

Yes; for three years, as a Teaching Assistant, and then I left at the end of three years. I broke my ankle in February of that year, and that I think upset me more than a broken ankle should upset a young girl, and I guess I just gave up on that. I came home and literally could not find a job teaching math. It’s hard to believe in this day and age when they’re crying for math teachers.

Abbate:

This would have been ‘52 or ‘53?

Sammet:

‘51. I left in the Spring, when the semester ended in ‘51. They were not giving exams in New York City for math teachers, because they didn’t need them; and although I was away, my father was involved in local school boards, so he was able to track and find out exactly. They were not giving exams for math teachers, so although I was fully qualified, there was no way I could get a license if they weren’t going to give an exam. I looked into teaching at private schools, and for one reason or another, either I couldn’t find a job or I couldn’t find one that was satisfactory.

I finally decided, “Well, if I can’t find something in Westchester and I can’t find something in Manhattan, I will try New Jersey.” I was living with my parents on the West Side of Manhattan, so going to New Jersey would have been just as easy as going to Long Island or to Westchester. So I went over to New Jersey to inquire about what they required, and I met with them, and they reviewed my credentials and so forth; and they said I was missing two courses that they insisted on. And I said, “What are they?” One of them was an education course, which we agreed I could take in summer school at Columbia or something, and I figured, “Well, that’s a waste of time and money, but it’s not very serious. Maybe I’ll learn something more.” They had one other course that they insisted that I take, in order to get a license in the State of New Jersey: a course in the history of the State of New Jersey!

Abbate:

To teach math?

Sammet:

To teach math! And I protested, politely, and then I screamed and yelled, and I said, “This is ridiculous: I am not teaching social studies, or geography, or anything where the history of the State of New Jersey could make any difference!” They refused, and so I just walked away, and looked for a job.

I went to work at the Metropolitan Life Insurance company. It would have been the Fall of 1951. I went to work at the Metropolitan Life Insurance company, in one of their departments that were grooming actuaries. Now, I didn’t really have any interest in being an actuary, but that was the department they put me in, and that was all right; and there were four or five of us who were math majors, which is what they wanted. There were a whole set of actuarial exams that they told us about taking, and so on; but my actual job there was to sit and read numbers to a high school girl, to check the dividend tables, which came out newly each year. This is pre-computing. Each year they published books about dividends and premiums and whatever, and from one year to the next they didn’t change very much, but they had to be retyped; and if they had to be retyped, they’d have to be checked. I spent much of my time reading back and forth with another girl, but only a high school graduate. The worst job I ever had. So I thought, “I’ve got to get out of here.”

So I went around, looking, and I went to Barnard, and applied to Barnard and to Columbia: Columbia in the Math Department, as a grad student; Barnard as a Teaching Assistant, to start in the Fall of 1952. The only reason I didn’t leave the Metropolitan after a couple of months there is because I didn’t want to change jobs twice within a year. Even at that young age, I thought that was not a good thing to do. So I stayed. Turns out to be very important that I stayed, for a reason that I will explain in a minute, but let me go back to the time period when I was at the University of Illinois.

At the time that I was there, I was in the Math Department, as a Teaching Assistant. I was a graduate student in the Math Department at the University of Illinois from 1948 to ‘51. Over in the Engineering Department somewhere, they were building what became known as the ILLIAC, one of the early electronic digital computers. I think it is impossible to describe the amount of contempt that I and my fellow mathematician graduate students had for that! We considered that sort of one step above garbage collecting. That was practical stuff; not only was it dull, we felt, but demeaning to pure mathematicians, and incredible and impossible to consider. So although it was going on while I was there, I and my fellow mathematician students made it a great point to ignore it, as much as we possibly could.

Abbate:

Did you have any contact with it? Did you see it or use it?

Sammet:

No—because if anybody ever said anything about it, I ran in the other direction. Seriously! “Contempt” is exactly the word that we had for that. We considered that a contemptible kind of thing to do. At various times they tried to recruit us—”recruit” is too strong a word, but they did try and get some of us interested, because intuitively they felt that maybe if they had a few mathematicians involved, maybe something would come out better. But none of us were going to have any part of it. I never saw it; I didn’t talk to any of the people doing it; I had nothing to do with it.

This becomes relevant now to what happened at the Metropolitan Life Insurance Company. After I had been there about six or seven months, my supervisor came to a group of us—this group of a half a dozen people that were math majors and supposedly being groomed to be actuaries—and said, “We are getting some punch-card equipment in the Metropolitan. Are any of you interested in going to classes to learn how to use the punch-card equipment?” I thought to myself, “Well, that’s got to be on the same level with the ILLIAC: like garbage collecting.” On the other hand, reading numbers back and forth was even lower than that! So I said, yes, I would go to these classes on punch-card equipment, and so did several of my colleagues. We went, and to my complete and utter astonishment, I loved it! We were taught how to draw wiring diagrams, which is how you handle a punch-card machine. I never did wire anything physically, but did all my homework, and studied the book extra at home, and thought it was absolutely fascinating. So of course as soon as we got through with the class, they sent us back to what we had been doing. I don’t remember quite why, but we never got anywhere nearer to the punch-card equipment than the class. I guess they took us to see them, but maybe they hadn’t come in yet or something.

As I indicated, I then left the Metropolitan to go teach at Barnard and spend another year in graduate school in the Math Department. I decided at the end of that year that whatever type of ability it needed to get a Ph.D. in mathematics, I didn’t have it. So although I had completed all the course work, I decided, “I’m not going to be able to finish this.” I had gotten a Master’s Degree at Illinois, but that was very easy. So I went to look for a job, and that’s when I went to Sperry Gyroscope.

Abbate:

What was your specialty in math?

Sammet:

Primarily abstract algebra: group theory and the more general area of abstract algebra.

Abbate:

Did that sort of come back later, when you were working on algebraic languages? Or was it not really related?

Sammet:

No, no. It showed up in a different way. That is to say, my interest in that did eventually lead to interest in how to use a computer to do theoretical mathematics. And it did turn out that some of the early work that was done using a computer to do formal mathematics was in group theory, because it is an intellectual activity that lends itself in part to the help that can be given by a computer.

Abbate:

Okay. All right, so we’re back to Sperry Gyroscope.

Sammet:

They hired me. I was looking for a job where they needed a mathematician, and I was hired to do a couple of things. I was hired to run an analog computer. An analog computer—because again, this is 1953, and while there were a couple of digital computers around, I didn’t know anything about them.

Abbate:

So this was like a differential analyzer?

Sammet:

Yes, yes! And that really involves differential equations, which is why they wanted a mathematician, but I didn’t find that particularly interesting. It wasn’t too bad, but it really required more engineering interest and knowledge than I had. Certainly more interest than I had, and somewhat more knowledge than I had, because you’re really dealing with the physical characteristics of the analog computer. It’s not digital. Digital is very easy: you have zeros and ones, and you go on from there. This had to do with voltages, and—I can’t even remember all of the things. I did my job competently and was considered a great employee and so forth, but it wasn’t very interesting to me.

So I talked to my boss, and I got involved—at least part-time—in doing mathematical analysis of torpedo spreads. I was in a department that had a contract with the Navy to do certain things having to do with submarines, and I was really doing math: equations and things of that kind, having to do with error rates in sending torpedoes out. If you’re sending them this way, what is the error rate likely to be on hitting your target? And so forth. That was kind of interesting. It had one of these classic, classic situations in which I wrote internal memos that I then was not allowed to read—because I had clearance for confidential material, and I was awaiting clearance for secret material, and what I wrote they classified as “secret,” so of course as soon as it got stamped “secret” I wasn’t allowed to see it. I learned to keep an extra copy in my desk, because those were equations, and they built on one another, so I needed the material for the next part of the assignment! [laughs.] But that’s one of those things that, as I understand, still goes on today.

Sperry, Speedac, teaching courses

Abbate:

So you were at Sperry from ’53 until ‘58?

Sammet:

Yes.

Abbate:

So other than the analog stuff, you weren’t using computers there?

Sammet:

On the contrary: that’s where I got started, at Sperry. And you will see in a minute why the Metropolitan experience becomes very relevant.

After I’d been there a couple of years . . . I was working on this submarine stuff, which I knew was never going to get me anywhere, because they weren’t ever going to allow a woman on a submarine—I mean, even to walk on a submarine at that time, they didn’t allow women, and I wasn’t interested in making a career of submarines anyhow. But I was there, and I was doing my work, and I wasn’t thinking too much about what was going to come ahead.

My boss’s boss, my manager, came over to me one day and said, “By the way, do you know we’re building a digital computer?” I said, “I’ve heard of it. I’m not sure I know what that means.” He said, “We are building a digital computer, because we can see that that is the wave of the future.” Because this was a company that made its living on contracts with the Federal Government, and in particular with the military, so they could see, in 1955, that that was going to be the wave of the future. He said, “We’re building this thing. Do you want to be our programmer?” I asked the obvious question: “What is a programmer?” And he said, “I don’t know, but I know we need one!” [laughs.] I looked at him, and I said, “Well, is this anything like working with punch-cards?” And he said, “I’m not sure, but I think it might be.” So I thought, “Well, that punch-card stuff was fascinating. I’m not too inspired with this submarine stuff. I should at least give this a try.” So I became the programmer. No books, no manuals, no instructions, no nothing—and engineers who somehow thought that machine was going to run itself and didn’t want a programmer. Fortunately, my boss, who was a very nice and very smart engineer, knew they needed one, even if he didn’t quite understand what that meant.

Abbate:

Who was that?

Sammet:

His name was Arthur Hauser. A very nice electrical engineer, very nice guy.

And that’s how and why I got started in the computer business. (The analog computer was of no interest whatsoever, and not the least bit relevant. If I had never worked on an analog computer, it wouldn’t have made any difference relative to the digital computer— except many years later, when people were developing languages for digital computers that in essence simulated analog computers. That was kind of interesting, and I think I have a couple in my book.) But that’s how I got started, and it was an amazing experience, because there wasn’t anybody I could go to, and I didn’t really know anything and had to teach myself.

The engineers were building this computer, which was about the capability of the IBM 650—although I’d never heard of a 650, of course, and only found out later that’s about what it was. It was a drum machine, where time was measured in milliseconds.

Abbate:

Which computer was that?

Sammet:

Nobody’s ever heard of it. It was called “Speedac,” for “Speedy digital automatic computer.” Nobody’s ever heard of it: it was one of a kind; they built it to learn. There was one paper published on it somewhere, and somewhere I have it in my files, but it was not being built to be sold or anything like that; it was built so the engineers would learn what it meant to deal with a digital computer—because they knew they were going to have to do something or other with that, to get new contracts with the military.

It had paper-tape input. It was a binary machine, which meant you had to do all kinds of input/output conversion from decimal to binary and binary to decimal, in terms of getting data in and out. The loading program —what, in today’s environment, would be called the “loader”—was four or five hundred instructions, but the question was: How do you get that into the machine? After all, here’s a blank machine; it has nothing in it. So the engineers and I finally decided that what I needed to do was to write the smallest loader that I could possibly write (smallest in terms of numbers of instructions) that would then bring in the other loading program. I didn’t have to do conversion of numbers or checks; all the kinds of things that were in the basic loader I didn’t have to do to bring that in, because the assumption was that it had no numbers and it was perfect. So it turned out the shortest program I could write was 20 words long, and it was a 16-bit machine, and what the engineers did was rig up a toggle switch: so I had to take the 20-line program, convert it to binary—16 bits per word—and use the toggle switch to get this onto the drum. The trouble was that the engineers assumed, “Well, she’ll just go do this.” But if I made a mistake, I had to go back and start all over again—there was no saving of what I had input up to that point—and I’m not exaggerating: it took me three days to do it! Because I’d get down to the last word and get an itch or something and make a mistake; so it took me three days to get that 20-line program installed on the drum. Then I brought in the main loader, which worked perfectly, and went on from there.

The other thing that was relevant to that experience is that the engineering group that was using that computer was using only that. Sperry Gyroscope, however, had a Univac I—which they used for the data processing aspect: payroll and all of that kind of stuff—which we didn’t really get access to, because there wasn’t enough time [available on the machine]. Over the period of the first year or two, we began to hire; my bosses did the hiring, but the people were reporting to me: I was the group leader. So we began to get some people in.

We ran what at that time was called an “open shop.” It’s a term not used anymore. At that time, throughout the industry, particularly companies like Sperry Gyroscope that were aeronautical—TRW or McDonnell Douglas; those companies that were doing scientific and engineering computation—they were either open-shop or closed-shop. The closed shop meant that there was a staff of programmers. The engineers came in and talked to the programmers, and the programmer wrote the program, ran it, produced some answers, debugged it, and of course had to get help from the engineers as they went along. It meant the programmer had to learn a lot about the engineering or the task it was that they were trying to solve. An “open shop” meant that the [computer] people were consultants, that the engineers had to learn how to program. We helped them: we had software like loaders and utility routines that they could use, and we had programs to generate sines and cosines and so on; so we developed what became known as “system software.” But the engineers had to learn how to program. We ran some classes for them, and when they got stuck or had a problem, they came to us.

But there were programs that were simply too large for the computer that we had; they just took too much time or too much memory. Around that time, Sperry Gyroscope, at the top, top level, was getting involved with Remington Rand, which then led to a merger [in 1955]. That meant we had some access to Remington Rand Univac, and lo and behold, that’s when I met Grace Hopper and her group. They had developed some scientific programs and some scientific languages that were a lot easier to use than the basic coding that we did on Speedac. That wasn’t even symbolic assembly code: it was absolute addresses, so that if you wrote “add the contents of memory position 321,” you had to write “321,” and God help you if you made a mistake there! As contrasted with what was developed later—not on Speedac, but on other machines—namely, symbolic assembly, where you could write, “add X,” and elsewhere X was in some position, but you didn’t have to worry about that.

That’s when I first got to meet Grace and a number of the people who worked for her. Then it turned out that, under some very unpleasant conditions, we were able to get access to their machines. Where they were working was a very famous (in those days) place called 19th and Allegheny in Philadelphia, which is where they were assembling the Univac I’s as well as having offices. Grace Hopper had one group of people, and John Mauchly had another, semi-competing group of people, and I got to know both groups. If we wanted to come down at night, we could use their machines. We learned how to use some of these advanced systems: one called “A-3,” and then another one called “AT-3,” renamed “MathMatic.” The AT-3 / MathMatic is conceptually equivalent to FORTRAN.

We would work during the day, and [then] we’d go into Manhattan, take a five o’clock train to Philadelphia, maybe eat a sandwich on the way, and we would work on these Univac I machines until around ten-thirty or eleven, and catch the last train back from Philadelphia to Manhattan.

Abbate:

Where was Sperry located?

Sammet:

Sperry Gyroscope was on Long Island; it was in Great Neck. It was just across the border from New York City, because I can remember driving 25 miles, still being within the City of New York, and across the street was the Sperry Gyroscope plant.

Abbate:

So you had a substantial commute, then. To get to Philadelphia, I mean.

Sammet:

To Philadelphia? Yes, well: At five o’clock—maybe we left at four; I don’t remember—we took a train. The train was about an hour and a half, or maybe an hour; I don’t remember. So we got maybe two or three hours worth of work, and then had to come back, and we slept in the next morning, and then came in. We did this many times, sometimes a couple of days a week, but it was the only way to get the work done. The engineers were writing these programs—we taught them how to write these A-3 and then AT-3 programs—but we took them down and ran them.

The problem was that these were machines being put together and not yet fit for being shipped out for service, so they didn’t run all that well. There were a couple of times when we were trying to get work we had done onto the small reels of magnetic tape to carry back: because the regular reels that went on the machine were very large, but then there were some small reels that could be used on something else. There were times when we weren’t sure we were going to be able to get the results of the night’s work onto adequate-sized reels of tape to take back with us—because we were allowed a tiny bit of time on the Univac that was at Sperry, so we could print out results and that kind of thing. We had a lot of trouble with those machines, because they weren’t fully finished.

But as I say, I got to know Grace fairly well, and a number of the people that worked for her; and I got to go to lots of Univac conferences, which was very nice, because they had wonderful refreshments—first time I ever tasted caviar. Those days are long gone!

Abbate:

Were there all women your own programming group, or was it mixed?

Sammet:

Mostly men.

Abbate:

Really! Working under you?

Sammet:

Yes, surprisingly enough. In fact, the only woman . . . It’s been said by many people that you need to be, or it helps to be, a mathematician to be a good programmer. I’m here to guarantee that’s false. I had working for me a man who had a Master’s degree in applied mathematics and was a terrible programmer. He knew a great deal of applied mathematics, and he helped the engineers enormously, but almost any program he wrote didn’t work! I mean more than just the normal debugging; lots of time spent trying to get it to work. On the other hand, I had a young woman who came in to work for us as a clerk. She was going to type these punch-cards, because at that time, you wrote the program on a piece of paper, and it had to be punched: either punched into cards, which we weren’t using, or typed into a Flexowriter, which was something that ran a paper tape. She did this stuff for the Flexowriter and typed the paper tapes for us, and she got interested in seeing what we were doing. At one point she came to me and she said, “Gee, I’d love to get some kind of feeling for what’s really happening. Is there any way I can learn?” So I helped her, and I think by that time a book had come out; and she eventually got qualified, as far as I was concerned, to be a programmer. She was reclassified from her clerk job and became a programmer—one of the best programmers I ever had working for me. But she was a home economics major in college! She had zero mathematical background, and that was one of the first questions she asked me: whether or not she had to be a mathematician to become a programmer. And I said, “No, absolutely not. You need to have the right kind of mind. I can’t tell whether you have the right kind of mind, but you don’t need the mathematical knowledge to do programming per se. Now, if you’re going to analyze the problem and you need to worry about equations, that’s a different matter; but for example, to do system software, you don’t need to know any mathematics.” And I have either had working for me, or known of, excellent programmers who were history majors; one was a musician. So people have come in with all kinds of backgrounds, in spite of the fact that many people think—or at least thought at that time—that you had to be a mathematician; you had to be a math major. It’s just not true.

I was at Sperry Gyroscope for five years, of which three were working on Speedac and some use of Univac, and in particular the higher-level languages that Grace Hopper was developing. The other thing: I don’t know whether information about the courses I taught is on your question list?

Abbate:

I know you taught them, but I don’t have dates.

Sammet:

Okay. This, I think, is also interesting, because I taught for two years what turned out to be, in retrospect, one of the earliest courses taught on programming. I don’t want to claim it was the earliest, because I’m not really sure, but it turned out that on Long Island there were thousands—literally thousands—of electrical engineers, because there was Grumman Aircraft, and Fairchild, and Republic, and Sperry Gyroscope. There was a small college there named Adelphi, and there was a man who headed the Applied Math Department at Adelphi, who was a very enterprising, entrepreneurial man, in the very best sense of the word. He was Chairman of the Department, and he thought, “There are these thousands of engineers here, and some of them probably would like to get Master’s Degrees in applied math.” A very sensible idea. So he decided to set up an evening program in the Applied Math Department. (Obviously it had to be in the evening, because the engineers were working all day.) So he set up this Applied Math Department partly in the evening, and it turned out my boss at that time, who was an engineer, was teaching there, and his boss was teaching. They were teaching engineering courses, but they had enough math content that it could satisfy the rules and regulations to put this in this Applied Math Department. One day my boss came to me and said, “Hey! How about setting up a programming course? They’ve never had one.” So, I thought, “Aha! That’s a great idea.” The course was to start in the fall of 1956.

Now, there were no books. There literally was not a book on programming in 1956; there were only manuals. No available computers at Adelphi, obviously. So I ended up teaching this course at a graduate level—that is, credit for a Master’s Degree—which now is probably taught in elementary school! It was two courses, two different semester courses. First one was teaching them what a computer was. After all, these people, most of them, didn’t know what a digital computer was: input, output, central memory, registers—all of that kind of stuff. The first semester, I used Univac, which I of course knew better, and got Univac manuals for them; and because by that time I knew some of the Univac people in New York, I got them to agree to give me three hours of computer time at the end of the semester. So the whole class and I traipsed into New York, and they had written some programs ahead of time, and we ran the programs for three hours and did what we could, and then resumed the class. I also taught a little bit of numerical analysis—which I knew about as much about, probably, as you do!—but I got a book and managed to work my way through three lectures on numerical analysis. But it was very elementary; aside from the numerical analysis, it’s something you could teach in elementary school today, and that’s probably what they do.

The second semester, I decided I would use the IBM 704—which I knew nothing about, but got a manual and learned. So I taught them how to use the IBM 704, and managed to persuade the IBM education people in New York City to give me computer time at the end of the semester. Of course, the difference between the Univac and the 704 is that the 704 is a binary machine and Univac was a decimal machine. That was sort of mind-boggling for the students, because the change from decimal to having to deal with binary—not that they had to write in binary; they didn’t, but nevertheless, to even have to worry about it in any way, shape, or form—was kind of traumatic.

The second year that I did this, from ‘57 to ‘58, life was considerably better, because Dan McCracken had written an elementary book on programming. I think he made up a mythical computer, and it was a book to teach programming, and so I was able to use that book. It was the only book, and it happened to be very good. I don’t know what I would have done if it had been a bad book, but it turned out that it was a very good book. I used that the first semester. It turned out that in late ‘56, FORTRAN was released; so the second semester, I taught the 704, but I also taught FORTRAN—by the typical teaching dint of staying a couple chapters ahead of the students, because I didn’t know any more about FORTRAN than they did! I just read faster and read ahead. And so I taught, for those two years, material that would be taught, if not in elementary school, surely in high school, nowadays; but this was very valid credit for a Master’s Degree. I mean, that was sort of pushing the envelope at that particular time.

Abbate:

It sounds like you had an opportunity to learn a lot of stuff yourself while you were doing this.

Sammet:

Not only an opportunity: a requirement. I mean, there wasn’t anybody to teach me. If I had a FORTRAN question, I didn’t really know the IBM people, and there was no way that I could look it up. I do remember, I think in the first year that I taught the 704, they had a symbolic assembly program, and one of the students asked me a very reasonable question: “How does this work?” I said, “I honestly don’t know, but if there’s any justice in the world, it has to work pretty much like this,” and so I sketched the outline of how a symbolic assembly program would be written. Later I went and somehow managed to find out I was correct! I mean, I had a very broad outline, but I had the concepts, which I made up in 30 seconds and wrote on the blackboard [laughs]—and that’s exactly what they had done. Obviously there were a whole lot of details I didn’t know anything about.

Yes, it was a very self-learning experience for all of us at that time. There were not classes, except what those of us who were the pioneers were teaching.

Abbate:

You must have had experience already teaching yourself. Like with the Speedac: did they just give you the instruction set and say, “Here, figure it out”?

Sammet:

Yes! That’s all. So any courses that were taught on Speedac, I taught—or, eventually, people who worked for me taught. But there was nobody to teach me. The engineers gave me the instruction list, and said, in essence, “Well, a program: you’ve got to have these instructions, and this is what the machine is going to do,” and that was the end of it. I learned by myself.

And there’s something that comes out of this, many, many years later, when I was working for IBM, and I probably was in my early ‘30’s at the time. I’m bringing this up now, although it’s out of sequence, because I think it’s relevant to the question you asked. I hired two young guys: one Princeton grad, the other was an MIT grad, both math majors and both top of their class. I hired them to work for me in Boston. When I hired them, I was very perturbed, because they were, at most, twenty-two; I was, roughly speaking, ten years older, and I thought, “These people know almost as much as I do about computing, and programming, and things like that.” I was very upset for a while, because I thought, “Aside from maturity, what the heck do I have over these young guys? And why do they know 75 percent of what I know when they’re just coming out of college?” I finally realized that things that I had developed, or been around when they were developed, which might take a year or two to develop in terms of concepts and details, by 1961 were well enough understood that they were being taught in college. So something that might have taken either me or somebody else a year or two years to really develop, flesh out, and know exactly what was going on, they took in a six-month course. So then I stopped worrying about it, and I figured, “Well, this is going to happen. Things that take us a while to develop, people are going to learn in a classroom, and that’s the way life goes.”

Abbate:

It seems like there’s also a difference between knowing the answer and knowing how you got to the answer, which is what you knew.

Sammet:

Well, it’s not knowing how to get to the answer, it’s developing how to get to the answer, and that’s a big difference. But you’re quite right about the answer. When I was at the University of Illinois as a Teaching Assistant in math, I had a friend who was a Teaching Assistant in economics, and she said, “You, Jean, have a much easier time grading papers than I do.” And I said, “Why?” She said, “Well, all you have to do is look at the answer and see if it’s right. I have to read these papers and work through them, and so on.” I said, “Look: If you grade papers properly for a math exam, you can’t just look at the answer. You have to see how they got it. First of all, if they got the wrong answer, they may have had everything correct, except in line one they added three plus four and got eight; but all of their development was correct, except for that stupid arithmetic mistake! Now, that person doesn’t deserve to be terribly penalized. Whereas there may be somebody else who might conceivably have come out with the right answer, if it was a numerical answer, having made enough mistakes in the process of what he was doing, or using the wrong methodology, and by luck came out with the right answer! And so if I grade the papers correctly, I can’t just look at the answer—the answer may be the least part of it.” It’s like when I took a physics course as an undergraduate in college. In the lab we had, you had to do all kinds of silly experiments to produce “known numbers.” That is to say, you know the heat of the something is .391; most of the time, if it was supposed to be .391, we’d be lucky if we got .31 or something like that. But there was one set of experiments, and there were two numbers, and to three digits I hit them both correctly, having done this experiment. I didn’t fudge it; it just came out right. I finally decided, “I’d better fudge the results, because he’s never going to believe that I got that answer honestly. I did, but a physics professor is not necessarily going to believe that that’s what happened.” So I fudged some slightly and came up with point-three-eight-nine, instead of three-nine-one, or something. But with math papers, you can’t just look at the answer.

Sylvania Electric, MOBIDIC

Abbate:

So, you were teaching these two courses until you left Sperry?

Sammet:

I left Sperry. I decided that I couldn’t do anything more at Sperry. I wanted to get into a job that was going to be more computer-oriented—where computing was the significant activity, rather than being simply a support function to the engineers. And I went looking in the newspaper, as any reasonable person would do. In 1958, first of all, there were separate men’s and women’s ads. I don’t know if you know that . . .

Abbate:

I’ve heard that. I don’t know when that stopped.

Sammet:

I don’t know when it stopped, but I guarantee it existed in 1958; and I assure you that under “Women’s” there was no such thing as engineer, mathematician, programmer, or anything like it. It was clerks, teachers, housekeepers, whatever. So I learned to look under the “Men’s Help Wanted,” but there there was no listing for “programmer.” The term really was not in use at the time, so I had to look under “engineering” and then look at the job description to see whether or not there was something that I thought I could do. I eventually ended up at Sylvania Electric Products, in charge of the software for a computer called MOBIDIC. I don’t know if you have that on your list or not.

Abbate:

I’ve heard of it. I don’t know anything about it.

Could I go back for a second? I’m curious about the job ads. It wouldn’t be under “mathematician”; it would be under “engineer”?

Sammet:

Yes. The programming, yes. I don’t remember whether there were jobs listed for mathematicians. I just don’t remember. But I do know that isn’t what I wanted; what I wanted was a programming job. Programming was a well-defined profession by that point, but the newspapers didn’t know it, or the personnel departments didn’t know how to write ads for programmers, or they didn’t want to.

Abbate:

So they had this concept that it should be under engineering.

Sammet:

Yes.

Abbate:

Interesting. All right, so you ended up at Sylvania [in the Boston area]. You were happy to move to Boston?

Sammet:

Yes.

Abbate:

Had you been there before?

Sammet:

No, but there were only a few cities that I was willing to move to: Boston being one of them, Washington being the other. I don’t like to move. I lived in Boston twenty years, and I’ve been here [in Washington] twenty years.

Abbate:

That’s right; because you ended up at IBM in Boston, so that wasn’t a physical move.

Sammet:

Yes; because I refused to leave Boston, so IBM created a job for me. I’m serious!

Abbate:

Well, we’ll get to that. So you ended up at Sylvania. So they had no problem, even though the job was listed under “Men,” in hiring you for it?

Sammet:

Correct.

Abbate:

That was the newspaper’s idea, not Sylvania’s?

Sammet:

I don’t know whether it was Sylvania’s or not; all I know is they hired me to head up the software for MOBIDIC. In fact, the ironic part is that I went up to interview for one job—I’ve forgotten what it was called, “Systems Analysis” maybe—and it did not take very long for the man who was interviewing me and I to mutually decide that this was not the proper match. I didn’t have the skills that he needed, and his job was not what I wanted. He then sent me over to see the man who was in charge of the software group, a man who is well-known, at least to the older generation; his name is Carl Hammer, a very fine guy who later became a close friend of mine. He thought I was great, so he hired me to head up this programming group, which had barely gotten started.

Abbate:

Now, what was MOBIDIC?

Sammet:

The Army Signal Corps had the concept of developing an upward-compatible family of computers, which were to exist in differing echelons of the Army. The largest one— which was called MOBIDIC, for “MOBIle DIgital Computer”—was to fit in a van. Now, in today’s environment [of much smaller computers], that sounds incredibly stupid; but the MOBIDIC was of a capability roughly equivalent to the IBM 709, which in its entirety would take up a room, so the idea of a computer that not only would fit in a van, but could be driven safely and then perform, was incredible. There were about three or four levels. I don’t think you want a full lecture on the Fieldata system; if you are interested in pursuing that, there was a paper in the Annals. The smallest one was to fit in a jeep. Now, when you carry a laptop around, that sounds silly; but in those days, when computers took up rooms like this, that was an absolutely pioneering concept.

Abbate:

And that would have been a minicomputer? Sort of refrigerator-sized, or smaller?

Sammet:

Smaller than a refrigerator, I think; maybe half the size of a refrigerator. I don’t remember. But this was a pioneering concept, which most of the Army has forgotten about. I’ve talked to people in the Army who don’t know about this—never heard of the Fieldata program.

MOBIDIC was the largest, and so the Signal Corps gave Sylvania a contract to build a couple of MOBIDICs: to build the physical computers and to develop the software that was to run on all of these computers. The other, smaller ones were being built at different places.

Abbate:

So it was going to be a family, like the IBM 360?

Sammet:

Yes! But it was an upward-compatible family before the 360 was even a gleam in anybody’s eye. It was an upward-compatible family. The machines smaller than MOBIDIC were being built by, I think: RCA; Philco, which eventually went out of existence; and even IBM was building one of them. But we were supposed to develop software that was going to run on all of them—providing that the thing was not time-dependent, because if it’s dependent on time, then obviously, the timing being different on all of these, it’s not going to run properly. The instructions were the same, but if you write a program that depends on something happening within one second, and you have a much slower machine, it’s not going to work; and everybody knew that. So we were developing that software, and that was lots of fun.

Abbate:

What kind of background did Sylvania have in software development, at that point?

Sammet:

None. [laughs.] Well, this was 1958! And they had not built a computer, either; but, from the point of view of the military, there weren’t any companies that had this kind of background, either in hardware or software—or very little—so all they could do was look at the previous things that Sylvania had done. They had built something called “UDOFT,” which was some kind of flight simulator—sort of like a Link Simulator, but different. I don’t remember any of the details, but they had a credible track record to award the engineering contract, and presumably Sylvania’s proposal convinced them that they could learn to write the software. But they didn’t have any track record: I was it! Well, actually there were a number of people that had been hired, and they knew something, because by ‘58, people had been able to learn some things. So I was coming in to take over part of a group and hire some other people and get that started. And that was lots of fun!

Abbate:

It sounds like it! So you developed . . .

Sammet:

The system software for MOBIDIC, to run on all these smaller computers as well as MOBIDIC.

Abbate:

So that had to be translated into a whole bunch of different instruction sets for the different computers?

Sammet:

No, no. An upward-compatible family of computers means that going from the bottom to the top, all the instructions are going to be there. It may or may not work from the top to the bottom—that is to say, there may be instructions in the top that aren’t in the smallest machine. I think they all were in the smallest machine; I don’t honestly remember. But that’s what you mean: the instructions are the same. They may be implemented differently. In fact, what’s interesting is that the programmers learned all kinds of cute tricks, which I then had to stop them from using; because when you implement an instruction, such as an “add” instruction, depending on how the engineers implement it, you may end up with certain things in certain registers that have absolutely nothing to do with the “add” instruction, and smart programmers are going to find that and use that. Well, when you come to another machine, the engineers implemented the “add” instruction differently, and you don’t have this spurious stuff sitting in that register. So a program that was using that, which worked on the large machine, would not work on the next one. Although the main instruction was identical, the programmers were using this side stuff that didn’t work. So I had to put my foot down and get them to stop doing that.

Abbate:

That’s always a problem with standardization: trying to keep people from taking shortcuts.

Sammet:

Yes, absolutely.

[DISC 2]

Sammet:

I was at Sylvania doing MOBIDIC.

Abbate:

You were doing MOBIDIC. And you were there about three years?

Sammet:

Yes.

Abbate:

‘58 to ‘61?

Sammet:

‘58 to ‘61, yes.

Abbate:

You were doing MOBIDIC the whole time?

COBOL

Sammet:

After I got involved in COBOL, which was in 1959, I then shifted into a staff position; because the COBOL was taking up so much of my time that I could not run the department to develop the software and serve on the COBOL Committee, which was meeting very frequently.

Abbate:

Tell me how you got involved in COBOL.

Sammet:

There was a meeting called by a man called Charlie Philips, who was the Director of something at the Pentagon. He called the meeting, and there were a lot of people attending the meeting. They were from computer manufacturers, a couple of university people . . .

[JS brings out documentation on the COBOL meetings.]

Sammet:

All right. Charlie Philips, who had the title of Director of the Data Systems Research Staff, Office of the Assistant Secretary of Defense for Comptroller, convened a meeting in the end of May, 1959. There were about 40 people there, from Government organizations, users and consultants, computer manufacturers; and I went representing Sylvania. The purpose of the meeting was to see about getting some kind of a common language.

Now, I don’t know how much detail you want on this, because it is so well documented it seems rather foolish to take up time reading to you, or recounting to you, what is documented in excruciating detail here. So if you have any more subtle, general questions, I would try and answer those rather than the factual ones, if that’s all right with you, because it’s all right here.

Abbate:

No, we don’t need to do the whole history of COBOL.

One thing I’m wondering: It seems like most of your work prior to that had been in scientific computing, and COBOL, of course, is a business language.

Sammet:

Well, yes and no. The work at Sylvania was in the systems software, because we were developing the system software: the sort routines, the assembly, the loaders; all of that kind of stuff. So I had moved from the scientific area—scientific and engineering computing, which I did at Sperry Gyroscope; you’re quite right—to the systems work, which I like much more, at Sylvania. And the reason I got involved in COBOL was that, while there were numerous mathematical languages available, on at least the major machines—not on MOBIDIC, but on the major machines—at that time, the only business-type language was Flow-matic, which had come from Grace Hopper’s group on the Univac; and we at Sylvania were thinking into whether we should develop a business-type language. It was just that simple: because the customer obviously needed one, and we didn’t need to do anything like FORTRAN, because FORTRAN existed, or could be made to exist. We were looking into this, and that’s the main reason [I got involved with COBOL]. I just got intrigued with the idea of developing this language. I had had some mild interest in programming languages, but not a very pronounced one until I got involved in the COBOL work and just got very interested in this idea of developing a programming language.

Abbate:

Had you done any business applications?

Sammet:

No! [laughs.]

Abbate:

Of course, I’m not sure if the committee in general had a lot of experience in business applications.

Sammet:

The answer is “no!” We were all—almost all—systems software people or people with scientific-type programming. You’re right: we didn’t know much about business programming.

Abbate:

So was that a challenge? Did you try and talk to people more on the business side, to get some sense of what they wanted? Or did it seem commonsensical?

Sammet:

Most of it seemed commonsensical. The major inputs we had told us a great deal, because the experience with Flow-Matic that Grace Hopper and her group had had gave us a lot of insight. Flow-Matic itself, as far as I’m concerned, is the major input to COBOL, and I’m on record as having said if there were no Flow-Matic, there never would have been a COBOL. That seemed to tell us all we needed to know; and IBM was producing preliminary specifications for a commercial translator, which had some more advanced ideas—as one would expect, because they were being done several years later. So I think it was sort of a matter of common sense—unlike scientific programming, where if somebody says to you, “You have to develop something that’ll help you compute a trajectory,” your common sense is probably not going to tell you how you compute a trajectory. It might tell you, but it sure as heck does not tell me! I don’t know how to compute a trajectory. But if I know I have to compute a payroll, my common sense tells me what I have to do to compute a payroll, or to sort; I know I have to sort things. So while the business programming is much more difficult because of the concept of files and data, which you don’t have in the scientific programming, the ideas of what you have to do for the business programming, in my opinion, are much simpler. So common sense served the purpose, primarily.

The main challenge we had was to develop a computer-independent language, which was a crucial aspect of this. That is to say, it would have to run satisfactorily on “all machines”—or at least all machines that were big enough, whatever that meant. That was very, very difficult, technically, with regard to the data aspect of it. The procedures are not hard, but the data is: to get that to be machine-independent is very difficult.

Abbate:

Maybe this is all documented, but I’m wondering: Was that the biggest challenge, conceptually—how to make it machine-independent?

Sammet:

Well, that and the fact that everybody had a different idea as to what was the best way to do anything. I think this is very relevant to your question: When I was the Chairman of what was called the Statement Language Subcommittee—we were the people who were supposed to come up with the commands, as contrasted with the subcommittee that was working on the data and the data description—I decided, since we were all a group of strangers and didn’t know each other, I’d start with the simplest thing I could think of and see if we could get agreement. That was going to be the “stop” command. When the program ends, you’ve got to signify “end,” and I thought that that was going to be about as simple as you could get. It was not! I suggested “stop,” somebody else wanted “end,” and there were great debates about the subtle differences between those two words. Then it turned out that some people really wanted the machine to come to a grinding halt, and others wanted it to continue and revert control back to a mythical “operating system,” which some computers at that time had and some did not; and then there were some other people who wanted to be able to shift to another program, not just the operating system. So this allegedly simple, trivial idea of just signifying the end of the program was not at all simple. Because people had different ideas, that’s the first thing; secondly, they were coming from different computing environments, and they wanted something that was going to mesh nicely into what they had. If they had an operating system, they wanted to damn well make sure that the control reverted to the operating system. Those of us who were on machines that didn’t have an operating system wanted the thing to come to a grinding halt, so we could do whatever the next thing was. And these are non-trivial issues to work through, particularly among a group of fairly competitive strangers—because most of the people on the committee, at that point, were from computer manufacturers. The users and the more neutral people came in about a year later; but the basic committee was primarily consisting of computer manufacturers, who were obviously competitive.

Abbate:

What was it like trying to develop a language by committee?

Sammet:

Very difficult!

Abbate:

Did you divide it up? Or did you have agree on everything? How did that work?

Sammet:

In the last analysis, we voted. We divided the work up initially into two major sub-units and a couple of small ones. The two major subcommittees were a Procedure Language/Statement Language, which I headed, and a Data Description, which was headed by a woman named Mary Hawes. It turns out, interestingly enough, that on this committee were lots of women—and strong women: myself; Mary Hawes from Burroughs; Gert Tierney from IBM; Betty Holberton and Nora Taylor from the David Taylor Model Basin—and Betty Holberton was the most senior person, because she had worked on ENIAC.

Abbate:

Right.

Sammet:

It’s too bad you can’t interview her.

So there were a lot of strong women on that committee, and the Chairman, Joe Wegstein, really didn’t do anything. I mean, he was more of a hindrance than a help.

The Statement Language Committee moved ahead, and we got a number of things done, and in the committee we voted. I don’t remember whether there’s anything in here with things about votes, but we voted. For quite a while, unfortunately, the Data Description Committee didn’t get anything done. As we were developing statements in the language, we had to make certain assumptions; and when the full committee met, it was kind of a mish-mosh, because the Chairman was not running it very well, and half the work wasn’t even there in a preliminary stage to try and mesh with the other half, which we had done. Finally (and this is well-documented), a group of us decided the number of people was too large—by that time there were a lot of people coming in and out—and we would sort of volunteer to go off and do this ourselves. So we got the Chairman to appoint six of us—two from Sylvania, two from IBM, and two from RCA—that were going to go work as a group, and try and create a set of specs that the committee could work on.

The language had four parts to it: the statement language and the data description, which were the two major ones, and the data description was to be as independent as possible; the environment description, which was definitely not computer-independent, but where what you wrote was to be common among all of them; and the procedure division, which was supposed to be a hundred percent computer-independent. One of the interesting things, when we finally got stuff running, was the test that was run between RCA and Remington-Rand, where they wrote the same program and then ran it with the two compilers. That’s described in the paper and elsewhere; you can read about that if it’s important.

The six of us went and spent two weeks holed up in a hotel in New York and came back with a set of preliminary specifications, which we took to the full committee—which voted on them! I don’t remember exactly how many they accepted, but my recollection is that it was probably about eighty or ninety percent of what we did. The full committee had too many people to do anything this difficult, which had not been defined. Well, the only other attempt at defining a programming language by committee was the ALGOL-58 attempt. (That is documented elsewhere, particularly in the HOPL book. There are two papers in here on ALGOL.) In 1958 they were ahead of us, and ALGOL-60, of course, came afterward, because this was all being done in ‘59.

Abbate:

It sounds like at the end it was just a small group actually hammering it out.

Sammet:

That is correct; but the full committee either approved it or made some changes. But yes, you’re right: There is a limit to the number of people that can sit around a room and come up with a result. A larger group can vote—and we voted, and I’m not ashamed of that. Some people say, “Well, that’s terrible. You can’t define a language, or design a language, by a committee.” And I say, “Why not?” “Well, how do you decide?” I say, “We vote!” And they throw up their hands in horror.

Now, I don’t know whether voting is worse than a dictator—because you have one choice or the other. There are very, very few languages that are designed by one or two people; there are a few, but not very many. If you talk about a team, which is what FORTRAN was, or Ada: those were languages that had a lot of people involved, so you either have a team or a committee. If you have a team, you have a head, and the head is a dictator: you do it this way or that way. I’m not here to argue that a committee is the best way to do it, but I would defy somebody to say that a single person is the best way to do it, even if he has lots of people presenting him with ideas. It’s a debatable issue. You will find people who will say that if you have a single person making the final decisions (not all of the preliminary work, because it’s just too big a job), that it’s a more cohesive, organized whole. And I would counter-argue that if you have a group of people and you can get a substantial agreement among them (even if some disagree), that you can get just as coherent a whole. And the team leader, if he took a vote from his team, would surely have disagreements—because otherwise, why does he have to decide? Unless he’s going to put his foot down and say, “No, I reject the opinion of however many of you there are”—five or ten, or however many; if the team leader says, “I reject that; I’m going to do it this way.” Then you have a dictator. But otherwise, you have somebody who’s, in essence, deciding between two competing subgroups within his team. I’m not prepared to say that one or the other is better. I mean, I don’t mind somebody saying that they think the team leader approach is better; what I object to is when they say the committee approach is no good.

Abbate:

Are you happy with the way COBOL came out?

Sammet:

That’s an interesting question, which nobody has ever asked me before.

With one exception, yes. I was satisfied with our preliminary version—the so-called COBOL-60, the first version that got issued, disseminated by the Department of Defense. There were some facilities in there that were like a macro, and I had fought very hard to keep them in, so I was pleased that they were in. They stayed in through a couple of versions—I don’t remember exactly how many, but a couple of versions—and then the committee, which I was long since disinvolved with, took them out, on the grounds that nobody had implemented them. That was a fair conclusion. By that time, however, there was the beginning of a software industry, and the next thing I knew, some of these burgeoning software companies thought, “Gee, that’s a great idea!” And so they developed software—which they sold—to provide this facility that had been in the language for years, except that none of the manufacturers had implemented it. That’s the thing that I regret the most, and that I thought was very ironic.

Abbate:

Because you could have had it for free.

Sammet:

Exactly! Well put. And that’s exactly the phrase I have used in other contexts.

Abbate:

It was ahead of its time, I guess.

Sammet:

Yes.

Abbate:

How quickly was COBOL adopted? Were people dying to have it as soon as it was ready?

Sammet:

No. [laughs.] Well, the biggest problem came from IBM. IBM in some sense had to participate in the COBOL committee, because they couldn’t afford not to participate; but they had a competing language called “Commercial Translator,” which they implemented and were trying to give away—not sell, because software wasn’t sold then, except from these very few small manufacturers. So while they implemented COBOL, they also badmouthed it, and they were not very happy about it. But every manufacturer had to implement it, essentially because the Department of Defense issued an edict that said, “We won’t buy commercial computers from you unless you have a COBOL compiler.” So smaller companies like RCA, which had had a previous attempt at doing something; and Sylvania, which hadn’t even gotten that far; and some of the other companies like Burroughs, and so forth: they had to support COBOL, and they couldn’t afford to have a competing product. IBM did for a while, but people just started using COBOL.

One of the things that happened that helped, surprisingly enough, was the creation of the System 360 from IBM, which was announced in 1964. At that time, IBM still had both Commercial Translator and COBOL. Prior to that, on the 7080 and the 7070 and the machines that IBM used for business data processing, they had been using a system whose name escapes me at the moment. It was really a three-address pseudo-code, not a full-fledged programming language; they did not have anything comparable to Flow-Matic, for example. When IBM went to the System 360, they dropped support of that: they simply did not implement it on the System 360. So the people doing the business data processing programming had to either go backwards to symbolic assembly language or upgrade to COBOL, because the thing that they had been using wasn’t there. So they either had to go down to something closer to the machine, or up—and the choice was fairly obvious. They did have a choice between Commercial Translator and COBOL, but my vague recollection is that they probably didn’t know the difference, and so they figured they’d go with the standard one, rather than the proprietary one.

Abbate:

Now, you yourself ended up at IBM in . . .

Sammet:

1961.

IBM, FORMAC

Abbate:

It almost seems a bit ironic, given that you had just been working on COBOL and they didn’t seem to be the most enthusiastic adopter of that. How did you end up at IBM?

Sammet:

Well, after three years at Sylvania . . . Sylvania had gleams in their eyes of going into the commercial computer business: they even defined, and I think designed, a machine based on MOBIDIC that they were going to sell commercially. I thought that was fine, but then they began to drop that idea, and I thought, “Well, there isn’t much more that I can do here.” I don’t like hopping jobs, but three years seemed not unreasonable, and so I just started looking. I looked in a number of places, including IBM, and eventually decided to go there.

They created [a whole new group when they hired me]. It’s interesting: IBM is known for changing managers fairly frequently, and I had three managers before I even accepted the offer—because the offer was made to me by Person X; and he was transferred and replaced by Person Y, which is understandable; and then they did some reorganization, so the original organization didn’t even exist anymore, and there was Person Z, and it was deemed that I should be part of his organization. They were very nice to me: the original man who made the offer decided (and got permission) to set up an entirely new group in the Boston area, because I said, “I’m willing to talk to IBM, but I am not going to go to Poughkeepsie,” which is where most of the programming was. I had no interest in Poughkeepsie; and I did not, at that time, want to go to Washington, which is where the Federal Systems Division was, which had skillions of programmers because they were doing lots of work for the Federal Government. So I just said to IBM, “Gee, I don’t think you have anything for me that I’m interested in,” because those were the locations. I almost canceled the last interview with him. I think I had three interviews with him, and I was coming home from a conference and had made an appointment to see him, and I thought, “I’m tired; why should I bother? I have other offers, which are perfectly fine offers. Why should I bother going to see him again? A very nice man, but we’re not going to get anywhere.” But I decided, “All right, I’ll go.” And I’m very glad I did, because he said, “I’ve come up with a great idea. How would like to form a new group to do advanced programming—to do applied research and advanced development—in the Boston area?” And I said, “Why? Why do you want to offer me this position?” Because it was creating a brand new group, in a place where IBM didn’t have anything except sales offices. They had lots of programming people at Poughkeepsie, and in New York, and Westchester (which I wasn’t going to go to either). He said, “Because we think there are a significant number of programming people whom IBM would like to hire, who are like you: they won’t go to Poughkeepsie or New York or Washington, but they might like to stay in Boston or come to Boston. We’d like to take advantage of that. We can set up this kind of a group, and you’ll report in to me down in Westchester” (or White Plains, or wherever he was.) I said, “Gee, that’s worth thinking about.” Before I had a chance to decide [laughs], came these other two managers—and so I accepted, talking to the third one on the phone without ever having met him! And that’s how I got to IBM.

Abbate:

So they created the Boston Programming Center for you to start recruiting people?

Sammet:

Yes. I mean, obviously they didn’t create it just for me, but it served their purpose of having a place to attract bright, advanced people in an area where they felt they could get them. These people wouldn’t go to Poughkeepsie, maybe didn’t want or weren’t qualified to go to the Research Center in Yorktown Heights, didn’t want to go to New York, but might be attracted to Boston. Because my experience has been that geography is often—not always, but often—more important than the job. That is to say, a person will reject a good job in a place he doesn’t want in order to take a slightly inferior job in a place he does want. Now, nobody’s going to take a job he doesn’t want in a place he likes, unless the person has to go there for family reasons, or something like that. But a lot of people are heavily influenced by the geography.

Abbate:

And they were probably thinking they could pick up MIT and Harvard graduates who were there already.

Sammet:

Exactly. Part of the mission that they set up was that we were supposed to establish and maintain informal contacts with both Harvard and MIT, to find out what was going on there—in a perfectly open and appropriate fashion, you know; not a question of spying, but just to have an intellectual presence. And that’s why it was designed and created as a group to do advanced work—it was called “Applied Research and Advanced Development.” In other words, we were not building products to be sent out the door, and we were not doing pure research; we were doing something that was to be in between. It might lead to a product, might not; but something that would have some new ideas in it.

Abbate:

Was that in Cambridge? Where was that?

Sammet:

Yes. We ended up, after a couple of years, in the very famous building 545 Technology Square, which is where Project MAC was, and where the key MIT computing group was. So I got lots of things done in the elevator—seriously! In fact, one of the things that happened, because it was a computer-heavy building—lots of MIT people, and us, and I think some other small group—I came in one morning, and there were two people waiting for the elevator. One of them I knew; I think he was an MIT person. He said to me, “How did that program work? I’m sure you’ve been working on that over the weekend.” And I said “Yes.” He said, “How did the program work?” And I said, “I think I got all of the bugs out.” And the third person, a complete stranger, said, “Gee! If you’ve got bugs, maybe you ought to call the exterminator!” [laughs.]

Abbate:

It couldn’t have been much of a computer person.

Sammet:

No, the third one was not; the third one was a stranger. It was a stranger, probably working for Fannie Farmer Candy or something. There were non-computer people in the building. The third person was definitely a non-computer person, who heard “bugs, building” and went into mental panic mode!

Abbate:

So, they put you in charge of this new group. How did you decide, or who decided, what kind of projects you were going to work on?

Sammet:

My boss suggested to me, “Why don’t you try doing something with symbolic manipulation?” That was the extent of the direction I had! So I came up with a couple of ideas, which had to do with comparison of existing systems—nothing very inspiring—and then somehow I came up with the idea of extending FORTRAN to allow for symbolic computation. Symbolic computation is in part the kind of thing you do in high school, where you multiply “A plus B” times “A minus B” and get “A squared minus B squared”—or formal differentiation, or integration, or any of those kinds of things. I thought, “Well, that’s certainly symbolic manipulation, and I wonder if he’ll appreciate that.” I sort of gave him about three sentences, and he said, “That’s a marvelous idea. Go ahead!” So it was one of these “Eureka!” things. You know, you get a flash, and it’s a good flash. You get some that are not so good, but that was, I think, just a good idea.

Abbate:

So that was FORMAC?

Sammet:

That was FORMAC.

There were some other things that we got involved in that are too hard to explain. I mean, I don’t think it’s worth taking the time.

Abbate:

Right. FORMAC is the Formula Manipulation Compiler?

Sammet:

Yes; it was “Formula Manipulation Compiler.” There was another meaning—that’s close enough, yes.

Abbate:

So this would have been the second language that you . . .

Sammet:

. . . that I was involved with the design of; yes.

Abbate:

How much did working on COBOL help you with FORMAC? Or was it a completely different experience?

Sammet:

Well, the experience was very valuable, because I knew what it meant to design a language. The main differences were that in the case of FORMAC, we were adding on to FORTRAN, and that gave us certain constraints from a stylistic point of view.

Abbate:

So was it a superset of FORTRAN?

Sammet:

Yes. The basic concept, I thought, was a very good one—which most of the succeeding language designers in this area have not taken, have simply not accepted as being important enough. The basic concept was that FORTRAN does lots of scientific computation; we now want to add this algebraic manipulation, for which people spent literally weeks and months—and, in the 1800s, years—developing formulas. I could remember that from the time that I was at Sperry Gyroscope, where I would spend days or maybe a week working out equations by hand! At Sperry, when I was doing that, it never occurred to me that you could use a computer for it. I mean, this was before I knew what a computer was. FORTRAN provided a whole lot of the capability that you already needed. After all, in any language—for either business data processing or anything in the scientific and engineering field—you need things having to do with arithmetic; you need input/output; you need control structures; and you need comparison: you need “go to,” and you need an “if/then,” and all kinds of things like that. These things are maybe stylistically different from one language to another—FORTRAN and COBOL are quite stylistically different—but conceptually are the same for any language in that genre. (There are other kinds of languages that are very different.) Since FORTRAN had a lot of this that we would need—it had the control structures, the arithmetic, the input/output, and so forth—my thought was, “Let’s add on what we want to do, because it’s the same set of people [scientific programmers] who are going to be doing this.” You know, it’s not going to be the guy developing payroll who cares about FORMAC.

So there were constraints, from a stylistic point of view, in terms of what kind of commands and data we would create; and there were detailed constraints, because we wanted to make sure that it was all going to fit together. I hope that answered your questions. If not, ask it in a different way.

Abbate:

I think it did.

How long did it take to develop FORMAC?

Sammet:

Well, we started pretty much in the Spring of ‘62, and we had a running system probably at the end of ‘63, something like that.

Abbate:

So a couple years; a year and a half.

Sammet:

Something like that.

I didn’t bring the second HOPL Conference [volume]; I had a paper in there; but maybe I can look this up while I answer something else.

Abbate:

You were managing a team of a few people?

Sammet:

Well, in the entire group I had maybe ten to fifteen people. About six, seven, or eight of them were working on FORMAC, and the others were working on other projects. I don’t remember exactly. And we had a couple of part-time students, who were very good.

Abbate:

How would you describe your management style?

Sammet:

I don’t!

Abbate:

[laughs.]

Sammet:

No, I’m serious! I am not a guru of management and management style.

Abbate:

Well, I don’t mean in a business sense or anything, but I’m curious how one manages a technical project like that. Were people very independent, and sort of reporting periodically, or was there a lot of coordination?

Sammet:

There was a fair amount of coordination; and after a while I had somebody working for me who was doing most of the day-to-day supervision, because I had lots of other responsibilities. I had the other people that were working on other projects, and I was running what in IBM terms was considered an off-site location: that is to say, I didn’t have a personnel staff; I didn’t have all of the supporting things that you have when you are in a large plant. All those people to whom I either reported or got help from were in Poughkeepsie or in White Plains—because my boss was in White Plains, and he was in a very large facility, but he himself was reporting in to Poughkeepsie. IBM has lots of people in strange places reporting in to places where they’re not physically located.

But I was more interested in FORMAC than I was in any of the other projects, and so I kept a very close eye, and as people went along I reviewed what they were doing and made suggestions as to changes. I don’t remember whether I insisted on any changes or not. That’s too long ago, and I don’t think that’s even in my HOPL paper; I honestly just don’t remember.

Abbate:

That’s fine. Did you ever read Soul of a new Machine?

Sammet:

I have the book; I’ve never had time to read it.

Abbate:

It describes an almost anarchic management style, which I would not necessarily hold up as an example, but it made me curious about ways to do it.

Sammet:

No, I would tend to keep a fairly close control of things, if that’s what you’re asking. That may not necessarily be the best way, but . . . [laughs.]

Abbate:

Well, you got it done. And you got the IBM Outstanding Contribution Award, and that was for FORMAC.

Sammet:

For FORMAC, yes.

Abbate:

So I take it they were very pleased with the results of that.

Sammet:

Yes! You were asking about dates: The first complete draft of the language specifications were done by December 1962. Please remember that for a long time I only had one person working on this, aside from me, because I was shifting around from space to space in the Cambridge area, for complex reasons; and was hiring people, and that always takes time; and trying to learn what the heck IBM was all about. Because what they did with me was extremely unusual—namely, they brought me in as a manager, with certain authority and responsibilities; and generally speaking, I found out later, IBM almost never hired anybody as a manager.

Abbate:

Who hadn’t worked their way up, you mean?

Sammet:

It wasn’t so much a question of working their way up, as not having experience in IBM. So they wanted somebody to understand what IBM was all about before they made them a manager. In fact, it became sort of ironic, because I came in at a Senior Programmer level, which was the highest programming level—“Senior Programming Manager,” I think they called it—and at some point, that called for me to get sent to some middle management school. That was all right, except that then, after my boss told me I was going to go, somebody in Personnel started checking the records and noticed that I had never been to the basic management school. They called me up and said, “You can’t go to senior management school.” I said, “Why not?” “Well, you haven’t been to the basic one.” And I said, “Well, what am I going to learn in the basic one?” “How you hire, and fire, and evaluate, and a whole lot of other things.” I said, “I’ve been doing this for two years now. If I don’t know it by now, it’s too late.” So finally the Personnel person got persuaded that I should go this middle management school.

But getting back to the dates: I joined IBM in October 1961, and it took a while—many months—before I came up with this idea, and I still had other things to worry about. Just the sheer logistics of where I was going to have an office, because there was no office for me. IBM had what they called a “Liaison Office” with MIT. MIT was a good and an important customer, so somewhere on the MIT campus was an office that IBM had, for people who were supporting the IBM installation; and a very nice guy who ran that let me have a desk there. But that wasn’t going to get me much besides a desk, so I had to go look for commercial space. I didn’t know the first thing about looking for commercial space! Fortunately, the manager of the Cambridge branch office was very nice and was a friend of my boss, who helped me, and so we went and looked in some places. I can remember, I went in to investigate the ladies’ room, and he went in to investigate the men’s room, and so on. [laughs.] So I found some commercial space, which we were in for about a year and a half, and then we moved into the Cambridge branch office for another year or so, and then eventually moved to 545 Technology Square. The plan always was, from the very beginning, that we would eventually move there; but the building didn’t exist, and so I had to find a space to habitate until that building was available.

We had the preliminary, first draft of the language specs in December ‘62. Remember I only started in October ‘61, so I didn’t get started on this ‘til the spring of 1962, and we went through a couple of versions of that; so that was December ‘62. April 64, the first complete version was successfully running, after extensive testing. Then in November of ‘64, we released it as what was called a “Type III” program. IBM had Types I, II, III, and IV. Type I programs were the major supported stuff, like the compilers, and IBSYS, and operating systems, and so on; and this was all free. Type II, which was sort of application-oriented, again were issued free, with support. Type III were really experimental, no-support programs. That is to say, if the group that developed it wanted to release it on a take-it-or-leave-it, take-as-is, caveat emptor (except no buying involved) basis, then that was all right. That’s what we did, as an experimental program; and lots of people used it—used it very successfully, wrote papers about their use of it.

Abbate:

So this was less than two years.

Sammet:

We got it running in a year and a half, and then there were enough things that we had to do, like a manual and other kinds of things, before we could release it. Obviously, we didn’t need the manual internally, but we certainly needed one for customers. When I saw the manual for the first time—one of my staff people had arranged for this—I thought I was going to faint; and then I decided, “If anybody can learn how to read the manual, they will have no problem learning to read FORMAC!” Because what had happened, in an attempt to save paper—because we ended up with something that was about three-quarters of an inch thick—what had been done was to reduce each page to half: put two logical pages on one physical side, and then print front and back. So on a physical piece of paper, you had four logical pages, and while they were all numbered, just figuring out which way to turn the darn thing—as I said, I figured if they could do that, then they could learn to use the system.

Abbate:

So, it was pretty widely used?

Sammet:

I won’t say “widely used”; I don’t know what “widely used” means. It was used by a significant number of people. But one thing that you may find amusing, for the record, is this: At that time, IBM marketing people were trying to get a handle on how you could estimate usage for a potential product. For hardware, by that time they knew how to go about going to customers and saying, “If we had this piece of hardware, would you be interested in buying it (or renting it, or using it, or whatever)?” While they didn’t have all the techniques worked out, they knew how to do that for hardware. They had never done it for software, and so they were starting to wonder, “How can we get a forecast of what the usage for software might be?” Because, although this was still early in the development of software, people were beginning to get the feeling this was going to be important. So one of them came to me and said, “Would you be willing to be a guinea pig, and get a forecast for the usage of FORMAC?” And I said, “Sure; that sounds interesting. Here’s a list of the people and organizations that are using it.” We had been in contact with them because, although theoretically we could put the program out and not answer any questions, we obviously wanted to get feedback, improve it, and all that other good stuff; so any time that anybody called us or wrote us, we jumped immediately to try and help them. I gave him the list of people and organizations that were using it, and about two or three weeks later I got a call from, I think, the manager of the person I had been dealing with, who essentially accused me of being a cheat and a liar. He was really very rude to me, accusing me of all kinds of things, and I couldn’t quite figure out what the problem was. Finally, it turned out that , according to him, I had given him the list of people, and half the companies on the list weren’t using this. Why had I sent him on this wild goose chase? And I said, “Look, I have letters in my files—physical letters—from Joe Blow at the XYZ Corporation, using it.” Some of the letters said, “Gee, this is great. I’m sure you’ll be glad to hear that I did these great things.” Others said, “Gee, I had a terrible time with it.” But nevertheless, I had documented proof that this stuff was being used.

This manager was growling at me, and finally he sent his subordinate to talk to me—in fact, sent him up to Boston to look at the letters—and he couldn’t figure out either why there was this discrepancy. He said, “Nobody there has ever heard of this!” It turned out, for some of these companies that were using it—not all of them, but some of them—the engineers or the mathematicians who were using this were in essence bootlegging the time. They either couldn’t or didn’t want to try and get formal approval, so they would load this and run it third shift, and do all kinds of things to keep it as hidden as possible from their own management. And when IBM goes to do customer surveys, apparently—and I didn’t know this—they have a designated person in each company that is their contact, and that person then goes to investigate and find out whatever it is that IBM wants to know, if they want to help IBM. In this case they were perfectly willing to help, particularly since it took them no time at all to call up their operating staff and say, “Are you running this thing called FORMAC?” And the operating staff would say, “Never heard of it.” It was the formal contact who didn’t know anything about it. In some cases, even the head of the people running the machines, the operations group, didn’t know about it, because the guy who was using it was doing everything he could to make sure nobody knew about it! Once we found that out—I’ll give this guy who growled at me credit—he called up and apologized and said, “Now I understand; but we have some difficulties working with this.” I said, “I understand that. But really, you shouldn’t be accusing me of being dishonest, just because somebody else is hiding something!” So that was a fun experience.

Abbate:

What a funny story!

Something you said earlier made me wonder, when you talked about hiring and firing: What were you looking for when you hired people? You had said earlier that you didn’t think mathematics was necessarily the thing you had to look for. What did you look for?

Sammet:

Well, do you mean at IBM or Sylvania? There’s a difference.

Abbate:

I was thinking for IBM.

Sammet:

Okay, because at Sperry Gyroscope, I couldn’t look for anything except somebody I thought was smart. At Sylvania, I’d try to get somebody who’d had some exposure to computing. By the time I was at IBM, I was looking for people who were interested in system software—as contrasted with applications—who seemed to be smart and seemed to have a knowledge of what was the current technology at the time.

Abbate:

So at that point you could already be hiring experienced people.

Sammet:

Not necessarily experienced from work, but experienced from school. When I hired the people in 1962, these two particular graduating seniors, I don’t think they had computer science departments at MIT or Princeton, but there were plenty of computer courses for them to have taken. So I looked at their records, and then I grilled them about what they were supposed to know—to make sure they hadn’t just taken the course, but that they had learned something. Does that answer your question?

Abbate:

Yes.

Sammet:

I will tell you one thing that I think is kind of funny, and I’ve told this to audiences. When I was at Sylvania and trying to hire people to work on MOBIDIC: up till that time, most programming—of which there wasn’t very much—was application programming. You can understand why; the system software aspect of it was not as common. I would bring people in, and I would ask them whether they wanted to do useful or useless programming, which was a trick question to see how they were going to respond. Almost all of them decided that it was a trick question and they’d better say they wanted to do useful programming—at which point I would say, “Goodbye, I don’t want you.” I did this in a gentler way, so it didn’t scare them off. The distinction I was making—which I explained to them, and then we discussed—was that the so-called “useful” programming was the application program. You’re calculating a payroll, or you’re generating a list of something, or you’re computing a trajectory, or whatever. It’s the applications program. The thing I was facetiously referring as “useless” is the systems programming, because it doesn’t produce an answer. You never get an answer from a systems program; you just provide this so somebody else can get an answer. So that’s one thing. The other thing, interestingly, was debatable for quite a while, and some other people, in different companies at different times, had a different viewpoint than I did. In the early ‘60s, or even up until the mid-’60s, the question often arose as to whether, in the hiring situation, you preferred individual machine experience or programming type experience. In other words, did you care more whether the person had run a program on the 709, or whether he’d written a payroll? Whether he’d been on the RCA whatever, or had written a trajectory program? From the very beginning, I always took the position that I was more interested in the type of programming they had done, on the grounds that it was a heck of a lot easier to teach them a brand new computer than it was to teach them the type of programming we were going to do.

In fact, that concept had shown up as early as the mid-’50s, when large companies like insurance companies or engineering companies were looking to train people: they assumed there were no programmers around, so they were going to have to train people. The question then became, did you try and find somebody who had some minimal programming exposure, or somebody who knew about the kind of work you were doing? In particular, the insurance companies and the business data processing people very quickly found that it was a heck of a lot easier to teach about the computer than it was to teach about the application. If you don’t know anything about the insurance business, and you know all about a computer, the insurance business is not an encapsulated thing that you can teach. You can teach some of it, but a lot comes from experience. Whereas, if you’ve had experience in the insurance business—or a bank, another example—and you understand about how interest is calculated and things of that kind, the computer knowledge is encapsulated. I’m not saying it’s easy or short, but it’s encapsulated, so you can run a class. Now, maybe the class is six weeks, but at the end of it, the person knows, in theory at least, all he’s supposed to know. He needs experience and will make mistakes, but at least there’s a concrete body of knowledge; whereas for the application there is no concrete body of knowledge. You can’t teach all that you want somebody to know about banking, even in a six-week course. So I always was interested in people who had done systems programming. I didn’t care what machine they’d done it on; I figured they can learn the machine.

Abbate:

The machine might be obsolete in a couple of years, anyway.

Sammet:

Exactly: that was the other point.

Abbate:

It sounds like FORMAC was a big success.

Sammet:

Not that much of a success. I regret that it wasn’t that much of a success. People did not like the concept—future developers in the symbolic manipulation area did not go for the idea of adding on to an existing language; and because IBM never supported it fully, it eventually died out. Actually, there were people who kept it going long after IBM dropped any knowledge of it, because they felt it was useful, and it had an impact, but not as much of an impact as I had hoped.

Abbate:

Was that because IBM just didn’t think there was enough market for it?

Sammet:

I think that’s part of it. And they had a lot of other things that they were worrying about at the time. I do remember at one point, my boss and I went to a Division President to make a pitch to get more money to do more on this. My boss had been around forever and was well-known, but I was a young whippersnapper, figuratively speaking—not so young, but certainly a whippersnapper. So the Division President said to me, “Young lady, I’m going to tell you something that you’d better remember for the rest of your IBM career.” And I said, “Yes, sir.” He said, “We have, in IBM, more bright people and more good ideas than we can possibly afford to support, no matter how good the idea is. We have more of those than we can possibly develop and support.” That was a very important lesson to me. I didn’t like to hear it—[though] he did give us money—but it was a very important lesson, and it’s probably still true. I suspect it’s true of any large company, in almost any field. If it’s a company with a lot of bright people, there are probably more good ideas worth promulgating—if you had an infinite amount of money, you’d say, “It is worth spending money on this”—but they can’t support them all.

Abbate:

What did you work on after that?

Programming languages textbook

Sammet:

In 1965, IBM was very good to me, and they in essence let me write the book on programming languages. I don’t know whether you’ve ever seen this or not . . .

Abbate:

Programming Languages: History and Fundamentals.

Sammet:

Yes. You’ve done your homework very well; I congratulate you!

Abbate:

That was actually published in ‘69?

Sammet:

Correct.

Abbate:

So you were working on it for a few years, then.

Sammet:

Well, I turned the manuscript in in 1967. It didn’t take me four years to write it: it took me about a year and a half or two years to write it, but then it took them almost two years [to publish it]. This is long before the days of [electronic typesetting]. In fact, I tried to get them to use some advanced technology, and they didn’t want to; so we went through linotype, or whatever the machines they were using, in galley proofs and page proofs.

Abbate:

What made you want to do that?

Sammet:

Well, at that time, they were desperate for books [on computing], and somebody at Prentice-Hall got my name and kept hounding me—initially, to write a book on FORTRAN. And I said, “Why should I write a book on FORTRAN? I mean, I know it, but . . .” He said, “Well, what would you like to write a book on?” I said, “I wouldn’t!” He said, “Well if you had to write a book, what would you write on?” I said, “I’d write a book on COBOL.” He said “Fine. Go ahead!” I said, “I’m not interested in writing a book on COBOL.” He was a very nice guy, and we got to be friendly, and he would call me every few months and say, “Don’t you want to write a book?” And finally—I thought this would shut him up—I said, “Well, if were going to write a book, I’d really like to cover lots and lots and lots of programming languages.” He said “Fine! Go ahead.”

So I talked to my boss. I had been running this group for four years, and he thought I’d done a great job, and so he said, “If they want you to write this”— I had developed an outline and a prospectus—“we’ll give you the time to do it.” So they gave me full time, almost full time, to do that for a couple of years.

Now there are a couple of amusing stories related to the creation of that, and I don’t know whether you want to hear them or not.

Abbate:

Sure.

Sammet:

The only text facility that we had was 2741s or something, which was a typewriter-like device that produced material on (I guess) a reel of tape. I don’t remember the exact details. You could do some primitive text editing—word processing—with it; you could make corrections, move paragraphs around, and that kind of thing. So it had all the forerunners of the modern systems; [the software] just wasn’t as powerful, and the equipment wasn’t as powerful. We were using a system called ATS: “Advanced Terminal System” or something. I had some temporary typists in, because my secretary couldn’t do all of this, and I basically dictated, and had it transcribed, and then cleaned it up. I had an outline that had a preliminary list of languages, and I divided this up into chapters, and I thought, “Well, 300 pages sounds like a reasonable-sized book, and at ten chapters, that’s about 30 pages a chapter, and so that means 60 pages, double-spaced.” Whatever the ratio was, it was about 60 pages of double-space that was going to be generated for each chapter. I had gotten started and maybe written a chapter and a half and was getting things organized, and I got very depressed one day, and I thought, “I don’t want to write sixty pages on FORTRAN. I’m just not that interested in FORTRAN.” I was very depressed, because I thought, “What can I do? My boss has given me a very good deal, by any criteria, and I don’t want to go to him and say, ‘No, I’m going to drop this book.’” So I thought, “Well, I’ll keep plugging along and see how it goes.” I was working on other sections, and I probably didn’t come back to the FORTRAN for three or four or five months. By then I was in the swing of doing this, and I dictated the section on FORTRAN, and the typist comes back to me, and I look at it: fifty-nine pages! I had thought I had to write sixty pages on FORTRAN to get to my three hundred pages, [but] what I had forgotten was that it was going to be sixty pages per chapter—per chapter, not per language, and the chapter that FORTRAN also had ALGOL and a whole lot of other languages! And that’s how the book got to be eight hundred pages instead of three hundred. [laughs.] By the time I got into the swing of it, there was a lot more to say.

The other thing was that at that time, we were using the 2741 [with] some communications terminals over to Poughkeepsie, which is where they were actually running this thing. One of those summers we had a terrible time, because in Boston in the summer we were getting thunderstorms around three or four o’clock every afternoon, and we found out that the communications lines went down. Before we realized this, they lost a fair amount of material: it didn’t get transmitted. So we finally learned that at two o’clock, they’d better back up everything they had been doing, and then we’d go in very small increments.

One of the things that happened after it came out: Somebody came up to me and said, “Do you know you have the same paragraph in there twice?” I said, “No. I don’t know it, and I don’t believe it!” He said, “Well, I will tell you.” I don’t know what your reaction would be; mine assumed that what he meant was, I had a paragraph, and then I had duplicated it beneath there. That’s what I thought he meant. He calls and he gives me two pages. I said, “What do you mean, two pages?” He said, “Well, you’ve got this paragraph on this page, and the same paragraph on that page.” So, I said “Fine, I’ll go look it up, and I’ll get back to you.” I went and I looked, and by golly, he was correct! What had happened was this: I had written a paragraph that was involved with the history of ALGOL, and I had put it in the appropriate section. Then, as I was looking at the material later, still in manuscript draft form, I decided that it wasn’t entirely history—it was related to something else—and it belonged better in another section of that chapter. So I gave instructions to the typist to move it over there—and you can see what’s coming: she obviously simply copied it, didn’t move it. It was never caught, for two reasons. First of all, it was logical in both places; so neither I, nor the copy editor or proofreader at Prentice-Hall [noticed]. It made sense in both places, obviously. And they were sufficiently many pages apart that, if you’re dealing with galley proofs and page proofs, neither I nor the editor would spot it. And so, to this day, that same paragraph is in there twice, because on additional printings it was too big to take out.

The other thing that happened that was funny was this: I had gone through page proofs, and there were still corrections to be made on the page proofs, and being a suspicious cuss—my first book, my only book—I wanted to see the corrections to the page proofs. They told me I would have to come to Englewood Cliffs, New Jersey, to see them, because they were—there’s a particular term that means, in essence, “These are the pages ready for the printer.” Whatever term it was they told me, I said, “I don’t know what that is, and why do I have to come down?” Well, then they said, “Well, because those pages are ready for the printer, so, if you want to see them, you come down.” So I went down, and I am thankful to this day that I did, because there were about five errors in the page proofs. Four of them had been corrected. The fifth was the fact that, in the FORMAC section, all the verbs, which were supposed to be alphabetical in the page proofs, were not alphabetical. I had written all the instructions to make them alphabetical; they had not done it; and if I had not gone to see these pages, the language that I was directly responsible for would have been the one language in there that was all mixed up! And they had to fix them by literally razor-blading out and pasting stuff in, to get them to be correct. I made them send me a photocopy of that, because I said, “Having fooled me once, I am not about to . . .”

[DISC 3]

Abbate:

Okay, we’re back.

Sammet:

I have one story that I think is worth putting on the record. It occurred when I was back at Sperry Gyroscope. My boss had hired, and I was training, a young guy to do the programming on Speedac. After some preparation and so forth, I sent him up with his paper tape to go and run the program; I told him to telephone me if he had a problem, and I would come up. He went up there, and an hour went by; and while this [machine] was very slow, it was a small program, so I didn’t think it should have taken that long, and I was wondering if anything had happened to him. So after another fifteen, twenty minutes, I went up to see him. He was sitting there, with a big sheet of paper in front of him, with a great big grin on his face. I said “John, did your program work?”—which would be unusual, if it worked the first time when you were a trainee. He said, “No.” And I said “Well, why are you looking so happy?” And he said, “Do you mean they’re really paying me to have all this fun?” [laughs.] And I have used that story in certain talks, because to me, it’s always been the very best definition of programming that I know of, and people who really love programming think of it that way. Now, today, programming is a business and an occupation, and I don’t think people get as much fun out of it; but nevertheless, that has always been the best definition of programming that I ever heard.

Now we’ll go back to your questions.

Abbate:

Just to go on with what you were doing. You were at IBM for twenty-seven years; so ‘61 to ‘88? Something like that?

Sammet:

Yes.

Abbate:

And I think we’ve only gotten to the late ‘60’s. I know you worked on Ada, but that was probably . . .

Sammet:

That was later.

Abbate:

You probably did stuff in between.

Sammet:

Yes. After the book was finished, I was primarily in staff positions. I was doing some general programming language work. By that time I was in the Federal Systems Division, and I was doing a lot of staff work, some having to do with programming languages, other having to do with planning. One of the things that I remember happening that was interesting: I was trying to find out what programming languages were being used in what parts of the Federal Systems Division—that’s the part of IBM that does work for the Federal Government.

Federal Government

Abbate:

Were you in Washington at that point?

Sammet:

I can’t remember whether it was before or after I moved to Washington. It was certainly the mid-to-late ‘70’s. I moved down in 1979, but what I’m not sure of is whether it was just after I moved down, or somewhat before. It was probably while I was still living in Boston.

Well, two things happened: One, I got a telephone call from one of the physical locations of the Federal Systems Division, who called me and said, “I understand you can help me with programming languages.” I said, “I hope so! That’s what I’m supposed to be doing.” He said, “Well, we have a requirement from our customer”—which was some part of the military—“that we use a higher-level language, probably FORTRAN, for this project, and we don’t want to do that; we want to use assembly language.” I said, “Why do you want to use assembly language?” He went through the whole litany of the stupid reasons that people gave, at that point, for wanting to use assembly language: they could be more efficient; this, that, and the other thing. And I said, “Look, you’ve come to the wrong person. There is no way I am going to help you use assembly language as contrasted with using a high-level language. It’s not cost-effective.” So he said, “Oh! Well I guess you’re not the right person.” I said, “No, but let me see if I can help you.” So then I tried to work with him, to show him that the things he was concerned about could be encapsulated to a relatively small area of the very large program. He could do ninety percent of it in FORTRAN without ever encountering the worries that he legitimately had, and he could do a small portion in assembly language, and that probably would help the customer. So I was able to help him. But it was ironic that he was calling me—the strongest advocate of using high-level languages—to help him figure out how to convince a customer to use assembly language! The other thing that happened that was related to that was [when] I went down to Houston. The Federal Systems Division facility in Houston had the major contract with NASA for the programming of both the ground control for Apollo [and the Apollo itself]. The computers [at ground control] were IBM, special-made, but the computers in the Apollo were also IBM-made. I went to see both groups, and as the scheduling turned out, I talked initially to the group that was programming the on-board computer, where obviously time is critical—to put it mildly—and space is at a premium. They were using, I think, JOVIAL; I don’t know if you’ve ever heard that, but it was a language that came from the Air Force. They were using either FORTRAN or JOVIAL, but definitely using a high-level language for about ninety-five percent of their programming. I said “Fine,” and I looked at it, and that was all very interesting. Fine. That was in the morning. In the afternoon, I go to see the people who are programming the ground control computers, and I find out they’re using assembly language. I said, “I can’t believe this! If it had been the other way around, I could understand that. When every single word makes a difference, I can understand that if you can save two words, that may be critical to what you’re doing—and that would be true in the on-board computer, but don’t tell me with the 7090s you’ve got sitting here . . .!” They gave me all of the hogwash reasons that people give when they don’t want to use a high-level language, and I said, “Look, I don’t have the authority to make you do it. I’m just surprised NASA isn’t making you do it, and why the heck don’t you get with it?” I was always amazed by that, because I thought, “This is just all backwards.”

Abbate:

So there was a lot of resistance to using high-level languages?

Sammet:

Oh, absolutely! For years, there was a lot of resistance to using high-level languages. One of the very great things about the first version of FORTRAN was that the compiler was efficient—because the early arguments were that Joe Blow Expert Programmer can write code that is smaller and faster than can be produced by a compiler, and therefore they don’t want to use a compiler. Now, in the early days it was a very macho thing: because it was macho to write in code, but not macho to write in this—bleah!—high-level language. Really. So there was an enormous amount of resistance. It went on for years. I remember talking to some group in some company that was writing a COBOL compiler—but they were writing it in assembly language! And I said, “Why?” They came up with all of the arguments that people use, and I said, “Well, the payroll people would give you the same arguments.” “Oh, but that’s payroll. We’re writing the compiler!” Although there were very good tools and languages around for writing compilers in high-level languages, they didn’t want to do it. So the systems people sometimes were the very last of the group to do this. Fortunately, one of the things that helped turn the tide had to do with MULTICS. Have you ever heard of MULTICS?

Abbate:

Yes.

Sammet:

MULTICS, of course, I’ll say for the record, was a time-sharing system developed at MIT, jointly with General Electric, I think, and Bell Labs. MULTICS was written using the high-level language PL-1, which was very gutsy for what they were doing, because if ever anything needed to be efficient and small, it’s a time-sharing system. But they used PL-1, and used assembly language, or machine language, in just a few very small places where they had to get direct access to machine facilities, which you do if you’re dealing with a time-sharing system; you’ve got to get access to timing registers and things of that kind. So there was a very small portion that was written in assembly language, and I think every bit of it was justifiable. They wrote a paper at an AFIPS conference, I think, in which they said that the benefit that they got from using PL-1, which enabled them to make improvements in the MULTICS system, came not from recoding sections but from the fact that they were able to cope with the logic of the program much more easily than if they had these hundreds of thousands of assembly-line programs. So some more people then said, “Well, if MULTICS can do this efficiently, we’ve lost most of our excuse!”—which I always felt was great.

Abbate:

And then Unix, which was written in C.

Sammet:

Yes. But MULTICS really showed the way.

Abbate:

It was a lot earlier.

Sammet:

That wasn’t the first time that had been done, but it was probably the most major thing that had been done, where an operating system, or something of that complexity, had been written in a high-level language. It was probably the most complex system that had been written using a high-level language, up till that point.

Abbate:

Did you have the impression that men were more resistant to using the high-level languages? You said it was kind of a macho attitude.

Sammet:

Well, in the beginning, most of the programmers were men. So I use that word . . . I don’t know. I have dealt with so many more men programmers than women, that I really don’t know.

One of the things I finally learned to do, when I was talking to general audiences, and particularly students—I was giving a lot of very general talks on programming languages early on: I had a chart that had a list of advantages and disadvantages [of using a high-level language]. And there are some disadvantages, including the fact that the compiler may not be quite as efficient as the very best programmer. But then I always would upset these students, or this general audience, by saying, “It is indeed true that the very best programmer probably will produce better code than the compiler will. However, there are not that many ‘very best programmers,’ and most of them are working on writing the compilers! So don’t give me this argument about your payroll program, or your trajectory, needing all that competence. It’s probably better in the compiler.”

Abbate:

Let alone if you want to maintain it after it’s written.

Sammet:

Of course.

Abbate:

I’m curious about Ada. I guess if you were in the Federal Systems Division, since Ada was required by—was it just the Defense Department, or the government in general that wanted Ada?

Sammet:

No, it was the Defense Department.

Abbate:

I guess that makes sense that you got involved with that. But what was your involvement with Ada?

Sammet:

I had a number of different involvements. I was not involved in the very earliest stage, when they were developing the requirements documents, but I came in around ‘77 or ‘78. My boss or somebody said, “Oh, there is this thing called DOD-1” (which is what it was called initially) “that obviously they’re going to inflict on us” (which was my boss’s opinion). “Go look into this and see what’s going on, and tell us what we ought to do.” So I went, found out what was going on, came back, organized some organizing committees in IBM, and so forth.

One of the interesting things that happened with regard to the name “Ada”: it was not my idea, but I was telephoned about it. Does the name Bill Whittaker mean anything to you? He was the original honcho for Ada. It was his idea to develop this single language for the military. I had met him, but I hadn’t really been much involved, and he called me one day and said, “Do you know of a programming language called ‘Ada’?” And I said, “Not off the top of my head, but let me look at my list of 1200 languages.” Then I said, “No, I don’t seem to have any language called ‘Ada,’ and I’ve never heard of one.” He said, “Good!” I said, “Why?” He said, “Because we want to use that as the name for DOD-1.” And I said, “Well, you can’t base that on me. I mean, I don’t know.” He said, “From what I have been told about your files, if you don’t know of a language called ‘Ada,’ there probably isn’t one!” [laughs.] True story! Absolutely true! It was not my idea to use it, but I was used to help confirm that there was not another language [called Ada].

Anyway, I was involved with IBM. I set up a coordinating group; I organized courses; then I was put on what was called, by the Department of Defense, the Distinguished Reviewers Committee, which involved looking at the specifications and doing all kinds of things relating to that, which I can’t entirely remember, but I spent a lot of time doing that. Then when SIGAda got formed—which was the Special Interest Group on Ada at ACM, previously called something else—I volunteered and headed up the Policy Committee for a while, to get appropriate speakers at the SIGAda meetings from the government sources. And I was doing a great deal of work in IBM: trying to get the Federal Systems Division educated; trying to get Project Managers to understand they had better get with this. It was my idea to develop a design language based on Ada, which was very tied-in with some other work we were doing in the Division, and so I was promulgating that; and that was pretty much full-time for a number of years. Oh, and I had a lot of fun writing a paper that ended up in the DOD package that went out from the Ada Project Office. I wrote a paper, which was published in the ACM Communications, called “Why Ada Is Not Just Another Programming Language.”

Abbate:

I saw the title. I was going to ask you about that.

Sammet:

Because, in my opinion, it was not. It was developed in a superb way, and I am on record as having said—I think I said it in the paper; if I didn’t, I certainly have said it orally—that they did an almost perfect job of getting that language created. I’m talking from a procedural point of view: I’m not saying the language is perfect; I’m saying that the procedures they used were darn close to perfect. I was glad to say that, and commend them, and so on; and that meant something to them, because of the reputation that I had with regard to programming languages. They simply had a set of procedures that, by golly or by gosh or by smarts or by luck or whatever, was as close to perfect as I think you could get, in terms of getting everybody involved in the right way, and at the right time, and in a fair way—as contrasted with the way most languages are chosen or designed. There’s nothing perfect about the way COBOL was created, I guarantee that!

Abbate:

So who was working on [Ada]?

Sammet:

In essence, if I can remember—I should have thought to reread this thing—first, they put out a series of requirements documents. Usually, when people design a language, they design the language and they say, “Oh yes, now it’s going to be used.” I mean, they have [only] a vague idea in their mind [of the user’s requirements]. But these were very detailed requirements documents about what kind of arithmetic was needed; whether real time was needed; and they had a whole slew of these things, starting out with something called “Straw Man.” First document that was put out was a very rough draft of what the requirements are for this language, which was going to be common across the military services. I don’t remember all the things: the kind of control structures that were needed, input/output, real-time, arithmetic—all kinds of things that were in there—and that was refined through a series of documents. There was Straw Man, Wooden Man . . .

Abbate:

[laughs.]

Sammet:

I’m not joking! I have the documents; I’m not joking. Let’s see: Straw Man, Wooden Man, Iron Man, Tin Man, Steel Man. I’m not sure that’s quite the right sequence, except Steel Man is the last one. Then, they put this out for competitive bid, for four companies to come in with some very preliminary language designs. Then they had those evaluated, but they color-coded the four proposals—Red, Blue, Yellow, and Green—and they stripped off the identification for each company, so that the reviewer theoretically did not know which company had produced “Yellow.” Now, that didn’t stay a secret all that long, but at least it was an honest attempt to have the reviewer choose based on the technical material, rather than on the fact that it came from Company XYZ whom he liked or didn’t like. They narrowed the four down to two, and had them submit pretty complete language designs. By that time, of course, it was no longer a secret; but they then had them evaluated, again in a fairly formal fashion. And then there was a DOD-type coordinating committee, which made the decision, and then they set up this Distinguished Reviewers Group to review everything that was happening, because these were still preliminary specs, and they had to be fleshed out and made final. They kept up this process of making sure that the community, however you want to define it, [had a chance to be involved]. I often said that anybody who didn’t comment on Ada, up till the time the final language specs were put out in the early ‘80’s: Anybody who didn’t comment up till that point is because either they were six feet underground somewhere, or they didn’t want to. I mean, anybody who wanted to comment was allowed to, and the comments were read and taken seriously. Presumably some were taken more seriously than others, but it was as fair a process as you could possibly imagine. So a lot of the reasons that Ada is unique was because of the process by which it was developed.

Another [reason] had to do with a number of the technical characteristics that are in there, many of which were unique at the time it was initially developed.

I was involved with the Distinguished Reviewers’ Committee. I did go over to Versailles when we had a meeting with the design team, because although the vendor of record was Honeywell-Bull, whose headquarters were in, I think, Minneapolis, the people doing the work were in France. It was a team that was part of Honeywell-Bull, but they were in France. They worked in Versailles, headed by a Frenchman named Jean Ichbiah. I got to know him fairly well, and we developed a routine that amused everybody any time we meet. I would say, “Hello, Jean” [pronounced American style] and he’d say “Hello, Jean!” [pronounced French style].

ACM

Abbate:

[laughs.] Now, that same time, in the ‘70’s, you also were very active in the ACM: the Association for Computing Machinery.

Sammet:

Oh, yes.

Abbate:

When did you first join?

Sammet:

I first joined ACM, probably in the late ‘50’s. I first became active in 1965. Do you want to know why I became active? Is that relevant?

Abbate:

Yes.

Sammet:

We were working on FORMAC. I knew there were other activities going on—there was important work being done by Al Perlis at Carnegie-Mellon, and at Bell Labs, and other work going on elsewhere—and I wanted to just chat with these people and exchange intellectual information. But working for IBM, you can’t just do that, particularly if they’re not with the universities; it’s hard enough with a university. It’s not that what we were doing was secret or confidential, but if you work for IBM, you just don’t go around chatting, in general, about everything that you’re doing. I thought, “Well gee, it would be nice if we could have a legitimate forum in which that could be done.” So I thought, “Aha! ACM has conferences, and sub-units, and all kinds of things. Maybe that’s a good venue, if we could get something set up.”

At that time they had what were called Special Interest Committees and Special Interest Groups, the first being just a way-stone on the way to a more permanent status as a Special Interest Group. So I wrote to the volunteer who was in charge of that, whom I knew, but not well, who happened to be a woman. I said, “Dear Jane Doe”—whatever her name was—”I think I might be interested in forming a Special Interest Committee on symbolic manipulation. Could you please tell me what’s involved? What do I have to do? I don’t know anything except that these things exist.” And I didn’t get an answer. I sent her another letter, didn’t get an answer, and because I’m a persistent kind person, I wrote to the President of ACM, who at that time was George Forsythe, a well-known numerical mathematician at Stanford. He was the President, and I knew him—not well, but I wrote and said, “Dear George, I can’t get an answer out of Jane Doe. I’m thinking maybe it would be useful to ACM to have a Special Interest Committee on symbolic and algebraic manipulation.” And I get back a letter from George Forsythe, “You are now the Chairman of the Special Interest Committee . . .” [laughs.] Things were a lot more informal then!

So I started to get this organized, and I thought the best way to flush out all of this work that’s going on was to hold a conference. So I organized a conference to be held in March or so of 1966, to be held down here [in Washington]. It turned out to be a rip-roaring success, in which all kinds of papers came up from underground that we didn’t know about. There were papers; there was a newsletter; and so it became quite legitimate to discuss—within reason—what we were doing with other people. We weren’t just in the position of saying, “Well what are you doing? But I can’t tell you what I’m doing.” There was a viable venue.

I guess maybe I did a good enough job of that, or something, and I was asked to run for what at that time was the Northeast Regional Representative on the ACM Council, and I won that election, and I was on the ACM Council for two years. Then in 1968, the President of ACM, who was at that time Bernie Galler, who has since become a very close friend of mine, appointed me the Volunteer Chairman of all the Special Interest Groups and Committees; so I had the responsibility of seeing that things were going well, helping to approve new ones, and so on. Then I was nominated for Vice President, and lost, in 1970; and I was elected the Chairman of SIGPLAN, fairly logically: it’s the Special Interest Group on Programming Languages. And then in 1972, I was elected Vice President, and then elected as President in 1974. ACM does not have an automatic accession: that is to say, the Vice President is not the President-Elect, or anything like that. It’s a brand-new nomination and a brand-new election.

So spent a lot of time on ACM activities. IBM was very generous in giving me quite a bit of time—because that can be very time-consuming.

Abbate:

It definitely can be.

Gender

So you ended up being the first woman to be President. Was that an issue at all?

Sammet:

For whom? I mean, it wasn’t an issue to me!

Abbate:

In your perception, gender was not an issue?

Sammet:

I was the first [female] Vice President, and then the first President. I must tell you, I am one of these women who simply has ignored this issue most of the time, and I’ve had a number of people who are very strong feminists who are annoyed at me, because I have not chosen to make an issue out of this all the way along. I have said publicly that in general, I was not discriminated against; and I think part of the reason is because I decided very early on that if I didn’t get something I wanted—a promotion, a raise, or an assignment, or whatever—that I was not going to tell myself that I was being discriminated against, unless I could really convince myself. It had nothing to do with anybody else. I just was not going to go into a mindset that said anytime I didn’t get anything I wanted, it was because I was a woman. So I just pretty much ignored this.

The only time that I can remember it coming up was a few months after I had become President. One of the women volunteers that I knew somewhat casually came up to me and said, “Gee, I’m so glad you’re President.” I said, “Thank you.” “Now that we’ve got a woman President, we can fix the ACM Constitution.” I said, “What’s wrong with the ACM Constitution?” And she said, “Well, it says ‘he’ and ‘him’ everywhere, and what we need to do is to go through and clean that up.” This is 1974; not an unreasonable viewpoint. I just looked at her, and I said, “There are a great many things that I would like to try and accomplish in my two-year term as ACM President. That doesn’t even appear at the bottom of my list; it’s in outer space! If you want to do this, and persuade the Council that they should do this, and so on, I will simply sit back, but I am not going to lift a finger to help with this. I have better things to do with my time, and the Council’s time, and staff, and so on.” I don’t think it’s ever been done. I’m not sure; I haven’t looked in a while.

That’s the only time that I can recall anybody making an issue; and that was another woman—it wasn’t even a man. Now, what the men said behind my back, I don’t know. I have asked a couple of them that have since become very, very close friends, and I get a little hemming and hawing, but the impression I get is that: Yeah, like anybody else, there was a lot of talking about me behind my back, but it had to do with characteristics that I possessed as an individual, and not because I was a woman. You know, if I was too this, or not enough that, they probably—at least, this is what my friends told me—they probably would have said the same thing about a man with those particular characteristics. Now, whether they were just shielding my feelings, or not, I can’t tell; but I think I was close enough to them that if there had been significant problems with my being a woman, I think they would have at least told me that much, even if they didn’t give me the details. So, I can’t guarantee that there wasn’t a lot; it wasn’t overt, is all I can say. I didn’t see it.

Abbate:

It’s been my impression, actually, that even from the beginning, ACM was quite open to women. I’ve looked at some of the early membership lists, and there seem to be a relatively substantial number. I don’t know if that makes it different from other computer societies.

Sammet:

I don’t know. But you see, we were helped by a couple [of things]. We were helped, indirectly but enormously, by Grace Hopper, who showed, “Here’s a competent woman who can do all kinds of great things.” The other thing—maybe I should have said this when I was back with Sperry, but it’s a point that can be made at any time: In the early days, in the ‘50’s, when programmers were an unknown breed of person, there was a theory in the minds of many of the men that women would make good programmers. One of these blanket generalities that, of course, is stupid; but the reason is funny. The reason that they used—and a couple of them said this to me—is that the engineers, in the early days, viewed programming as a clerical activity. After all, you’re writing a whole lot of stuff, and you’ve got to get every single character to be correct. If you have a single character or space that’s wrong, the program isn’t going to work! And they viewed that as a clerical activity. Maybe it had some intellect, but basically clerical. And who were the clerks? Clerks were all women. So they deduced, therefore, that women would make good programmers. Now, that’s, of course, sheer nonsense, and no more sensible than saying mathematicians ought to be good programmers. But that may be one reason that there has been, I think, less prejudice against women in the computer field than in the hard sciences like physics or chemistry or engineering. Chemistry, of course, had Marie Curie as a role model—but Grace Hopper was one; and Betty Holberton, although she was not as well-known, has to be the pioneer of all pioneers. After all, she worked on the ENIAC, and her name is on the Univac patent, so you can’t get much more important than that.

Abbate:

She seems to be legendary, from people I’ve talked to.

Sammet:

What I’m saying is that there are least two very significant role models for the women to look at. I was a contemporary, more or less, of Betty Holberton’s. Grace Hopper was, of course, considerably older. But frankly, I don’t think about the issue of women, and prejudice, and discrimination, and all that, except when somebody asks me. In fact, I remember, many many many years ago—probably 20 years ago—being interviewed by somebody for Computer World. I guess maybe he was interviewing me after I became President. I don’t remember what it was, but they were interviewing me, and they wanted me to go into a long harangue about discrimination against women, and I wouldn’t do it. I simply gave them my standard reply, which is that I refuse to think I’m discriminated against, unless I can convince myself that I really am. Just because I don’t get what I want—and just because a man got it—doesn’t mean that they gave it to him because he was a man.

In fact, I had more trouble hiring secretaries than programmers. When I was setting up the group in Boston for IBM, I was hiring programmers—men. I had to get a secretary. I’d had a series of temps, and I had an employment agency that was looking, and she called me one day and she said, “I think I have almost the perfect candidate for you.” And I said, “Well, what’s the background?” “A very good background.” I said, “What’s the ‘almost perfect’ about it?” “Well, she really would prefer to work for one of the men. Could you assign her to one of the men?” And I said, “Look, you know this is a one-secretary office. I’m in charge of the office. Ipso facto, she works for me. If she doesn’t want to work for a woman, don’t send her over.” And I learned, thereafter. If there was a male candidate for something, I never had Personnel ask him if he was willing to work for a woman; because I figured he knew I was a woman, and if he didn’t want to work for me, he wasn’t going to allow himself to be interviewed, or apply to me. The women, it was much trickier to have to do. I learned, at IBM and elsewhere—permanent and temporary—I learned to ask the Personnel people (I mean, before all the days of the rules when you can’t do these kinds of things!)—I learned to have them find out, either directly or subtly, whether they objected to working for a woman; because I don’t want to waste my time interviewing a potential secretary who doesn’t want to work for a woman. One of them, who turned out to be crackerjack and worked for me for three years, was asked that, at my request, by Personnel, when I was still up in Boston; and she said, “Am I willing to work for a woman? Well, it depends on the woman!” Which is exactly the proper answer! You know: Maybe I’ll like her and maybe I won’t.

So, I’m just not a flaming torch-bearer for these issues and these problems. I’m sorry. If that upsets you or your readers: tough!

Abbate:

Well, I don’t know who my readers are. It doesn’t upset me.

Was it your impression, in general, that IBM was a pretty equitable place to work, for women?

Sammet:

Oh yes! The only time we came close to having . . . Well, that’s a very good question, because it reminds me of something. I was running the group in Boston, and I had all these people working for me, including the group working on FORMAC. The Group Leader, the technical supervisor for these people, was a woman, who had transferred up from Poughkeepsie. It came time when I needed a manager reporting to me; I had too many people reporting to me. She was the obvious candidate, and the only other one who might have been a viable candidate for that position was a man who flatly refused to be a manager. He’d been in IBM; he’d been a manager; he refused to be a manager [again]. So she was the only candidate. I talked to my boss about this, and he knew how capable she was; he didn’t bat an eyelash. But I was a little nervous, and so I talked to a couple of friends of mine in IBM and a couple outside, and I said, “How are the men going to take it with two levels of women managers?” You know, a man might put up with one, but he might be a little unhappy with two. I talked to some friends, both in and outside of IBM. The people in IBM tended to know the woman, and knew me, and thought we could survive. One of the friends outside said, “Don’t do it; the guys will never stand for it.” I said, “I hear you.” One of the others said, “Well, you know, you’re taking a chance, but if you need to.” So we went ahead, and it worked fine. But then, as time went on, new people came in, and one of the new hires who came in was crackerjack, and she became a de facto technical supervisor. We never gave her an official position; I mean, we didn’t need to. There was one guy, who was obnoxious for other reasons, but he couldn’t take that. He could not take what were, in essence, three levels of women above him. At one point, he went to my manager and complained about something, without talking to any of the three of us. My manager told me, and he said, “Look, I can understand that,” and we decided how to handle him, and eventually the guy left. But he couldn’t take three levels of women, and I sympathized with him. I didn’t blame him; I could understand that.

[I didn’t just promote women] because I was looking to get the woman in the right place. I had enough problems on my hands; I would have happier, to be brutally honest, if the men had been more capable—it would have made my life easier, because I wouldn’t have had that problem! But the downright fact of it is, the women were the more capable for those particular slots, and I certainly wasn’t about to say, “Well gee, they can’t have those positions, because there could be too many women in supervisory positions!”

But IBM: I don’t know [how it was] at higher levels, but I didn’t see any problems, and there certainly were women who were put into very significant management positions, relatively early. I do know one story about IBM, having to do not with a woman, but with a black person. They were on a kick to get more black men—I mean, I’m sure they would have had women, but they were concentrating on men at the time—as salesmen. They were making some success, but apparently not in the South—for reasons you can understand, in the early ‘60’s, before the civil rights march and all of those kinds of things. Vince Learson—who was a Senior Vice President at the time and a holy terror, and eventually became President when Watson left—went down to one of the Branch Offices in Mississippi or Alabama, or somewhere like that. He went to the Branch Manager, and he said, “I understand you don’t have any black salesmen. Is that right?” And the Branch Manager said, “No, the customers won’t listen to them. We won’t sell anything.” Learson said, “Well, you have six months to hire, train, and have a successful black salesman.” The Branch Manager looked at him and said, “I can’t do it.” And Learson says, “Well, then we’ll keep putting in new Branch Managers until we find one who can!” As I heard the story, the Branch Manager managed to do it. He got a black salesman, trained him, took him around—as apparently they do with new salesmen anyhow—and got enough customers to be willing to deal with this black salesman. But there was the Senior Vice President, going down and saying, “You will do this, or you’ll have another job—a less preferable job!” [laughs.]

I don’t know of any cases about women, because I think the women problem was probably easier than black salesmen in the South. But one thing they did that I thought was fascinating: the position—I don’t know if they still call it that—of Customer Engineer, who was really the technician who went around to fix the equipment. I don’t know if they still call them “Customer Engineers” or not. You don’t know?

Abbate:

I’m not sure.

Sammet:

Doesn’t matter. They were called “Customer Engineers,” and they were the ones that went around in dark blue suits and a white shirt to fix the reproducer or the 7090 or whatever. But they were all men; and somebody must have said: “Why don’t you get some women?” The theory had been that the women can’t do this, because there’s a lot of equipment that has to be moved. So they issued job descriptions that said, “You must be able to walk up this many flights, and you must be able to carry this much weight, and you must be able to lift this and move it there.” These were all physical, honest-to-gosh physical parameters of the job; and anybody who looked at the description and knew what a C.E. did would know they were correct parameters for the job: you did have to be able to carry a big tool kit, and carry in whatever. It would be the equivalent, I guess, of carrying in a big TV set today, or carrying in a desktop machine, except that isn’t what they were carrying; they were carrying other stuff. The reason I’m very conscious of that is that I had a secretary when I was in Boston who was very interested in doing that. I don’t know why, because she had no technical background, but she was very interested, and she applied, and she got the job. She was a very strong woman, and she got the job. I always thought that was very interesting, because that’s a way of making sure you get people that are qualified without saying, “You have to be a man”—because there were obviously some men who could not do the job, and a great many women who couldn’t, but here was a fair way to describe it. I always thought that was great.

Abbate:

And that was in the ‘70’s, you think?

Sammet:

Let me think. I can tell you exactly when it was. It had to be in the late ‘70’s.

Abbate:

Let’s see. So you were doing Ada stuff probably to the late ‘70’s or early ‘80’s, maybe?

Sammet:

Oh, up until around probably ‘84 or 85. I was running a company-wide coordinating group, and a division-wide coordinating group, and doing all kinds of things; I can’t remember all the things I was doing. I do know I came within one week of being killed because of Ada! It had to be in the mid-’80’s: some madman, somebody who had been fired by IBM some time period before, went berserk. I don’t know how familiar you are with the area near Montgomery Mall [in Maryland]: there are lots of different kinds of commercial buildings there, about a mile or two away, and one of them was an IBM building that was called the “Rusty Bucket,” because it had a very rusty colored roof. Some madman went, drove his car in—which you could do at the time—right into the front lobby, because it was wide enough to do that, and took out a machine gun or a hand gun, and went around shooting people. I’m very conscious of that, for two reasons. One, I was in a meeting up at a building that was a quarter-mile away that was facing there—and we found out about this pretty quickly, because they got us away from the windows—and one of the men meeting in our meeting had a wife who was working in that building. But the main reason that I recall it so well is that, literally one week before, I had been running a corporate-wide meeting in the conference rooms that were right off the lobby—and he went into those conference rooms. It turned out nobody was in there; he went around to offices, and people were barricading themselves under desks and behind chairs, and I think he killed one person, and certainly wounded a number. But if he had come in one week earlier, or we’d had the meeting one week later, I might not be here.

Abbate:

I don’t know what to say about that.

Consulting, retirement

So you were just at IBM for a few more years after the end of that.

Sammet:

And you want to know what I was doing?

Abbate:

Yes.

Sammet:

They were very nice; they were giving me permission to rewrite this book, and so I spent most of my time for several years doing that, and also working on the first History of Programming—the first HOPL Conference—because I was both the General Chairman and the Program Chairman.

Abbate:

For the History of Programming Languages Conference.

Sammet:

Yes—the first one; there were two. And so that’s what I was doing the last few years I was there.

Abbate:

Did you leave IBM at the normal retirement time?

Sammet:

They were trying to clean out IBM, and a year or two earlier they had made a good offer, which I wasn’t interested in. Then they came up with another offer, which was very good, although I wasn’t particularly interested in retiring, because I was working on rewriting this, and I wanted to do that until I finished and then retire. And my boss in essence said to me, “Well, if you stay”—he put this much more gently, and it took more time, but the gist of it was: “If you stay, you’re going to have to go back to work and not work on the book, but if you take the retirement package, I’ll let you stay on the premises for another year. You can keep your office, your computer, a temporary secretary, all your files, and so forth.” That was not a hard decision to make, particularly because the severance package that they were giving us involved—it boiled down to essentially two years’ salary, I think. I’ve forgotten what the numbers were, but the net result was that it was two years’ salary, along with everything else. So I took it, and I stayed not one but two years, working on the book, which I did not finish then, and it became too big a job to do, and I finally left the IBM premises.

Abbate:

And then you became an independent consultant.

Sammet:

I did some consulting, and I occasionally still do some. Occasionally people call me because of the files I have—which I think you really would be interested in seeing, just to see what’s there. I mean, as a historian, and being involved with the computer field, I think you might be interested to see what’s there; so we could arrange some other time to go over.

Abbate:

Okay. So that takes us to the end of your computer career, that consulting part.

Sammet:

Yes, except that I’ve been doing varying things: I’ve been on some ACM committees; I was the Program Chairman for the second HOPL Conference; I was chairing, last year, the ACM Fellows Committee; and now I’ve got what is going to be an enormous job. I assume you know about the Software History Dictionary that CBI is doing?

History, final reflections

Abbate:

I know they’re doing some kind of software history project.

Sammet:

Well, you should better look up on the Web, rather than getting the details from me, but the gist if it is: It’s to be a software history dictionary, which means a list of terms—or names—with a history, going up to 1980. And I, logically enough, was asked to handle the programming languages thing. Bernie Galler is handling the operating systems; I know there’s a Graphics Committee, and I know there’s a Software Engineering Committee, and I realize I don’t know what the other committees are, and I should write and ask.

Abbate:

It’s a great idea.

Sammet:

But it’s a major task; because each committee, if I recall correctly, is supposed to have maybe a hundred to a hundred and fifty entries, and somewhere between 200 to 700 words per entry; so that’s a lot of entries and a lot of work to be done. For example, I would obviously write the COBOL entry, and have to compress it into whatever the number of words is. But then there are lots of other terms that have never had any historical description of any kind, and so it will be very good to get that.

Abbate:

That would be really helpful.

You have quite a perspective over that history. How would say the field of computing, or software, has changed over time since you started?

Sammet:

The changes are almost so enormous that I don’t know how to characterize them. It has changed from becoming something where only a few elite people were involved to something that the masses use, where a computer is a commodity like a VCR. There is, as far as I can see, almost no difference from a sales point of view, and almost from a usage point of view.

I will comment on the Internet, because I think the Internet is as significant—or is the most significant thing of its kind—since Gutenberg and the printing press. After all, Gutenberg and the printing press made material available to everybody—conceptually; I’m ignoring the practical difficulties, and you had to know to read, and not everybody did. But conceptually, you went from manuscripts, were the province of a few people who could write them and just a few more who could read them—because they weren’t physically available; if the monk was writing it in wherever, and you were living [somewhere else], you weren’t about to see this! But the printing press made material available to everybody in the world—minus a few practical difficulties, but conceptually, to everybody in the world. And the Internet, I think, has made essentially almost all information in the world that exists available to everybody in the world. And to me that’s the same kind of thing.

Abbate:

In terms of software—producing software, or designing languages—do you see major changes over time?

Sammet:

As best I can tell—and I’ve been out of that activity for a long time—as best I can tell, the production of software is no easier or better now than it was thirty-five years ago. That is to say, programs are produced, they are late, they are over budget, and they have bugs in them! And I can only say, “So what else is new?”

Abbate:

[laughs.] So there’s not been any success in rationalizing that, as far as you can tell?

Sammet:

My impression is that the number of errors per amount of code written is probably the same as it was forty years ago.

Abbate:

That’s an interesting metric.

Sammet:

That’s an impression. ... I am not in that part of this field any more, but some of the things that people were going to save the world with—proving that programs are correct, for example—have never panned out. Being able to write programs extremely quickly, accurately, and so forth, has never panned out. Things are better now than they were when we started, [but] I am not sure that in 2001, they’re better than in 1981, for example. I don’t know. I don’t know how one would measure that. That’s an intuitive feel; I may be all wrong—I’ve put as many disclaimers as I can put on this tape and this transcript!—but my impression is that software development is not much better done today than it was 20, 25, maybe 30 years ago.

Abbate:

It’s a good question.

What have you found to be the most satisfying aspects of working with computers?

Sammet:

They’re fun! I’m serious. When I left IBM premises, I had to get a computer at home, and I got the biggest thing I could get at the time, which was a 386, and I thought, “This is just great fun. It takes me back to the early days, when I was having fun!” I got my laptop only a few years ago, because the 386 served me very well, and the only thing that I wanted to do that I couldn’t do on the 386 was get access to the Web; I had email, but I couldn’t get onto the Web, so I had to get the laptop. I must say when I started with that, I hated it, because I was very good with DOS, and Windows 98 drove me wild. Everything I wanted to do could be done five different ways, and I had five different books, and depending on which book I opened, it gave me a different way of doing that thing; and it took me a long time to get used to Windows. Now I think it’s great fun, and I enjoy fiddling around. I confess, I enjoy playing with the Web. I really do! I like seeing what I can come up with in strange cases.

The strangest thing that I have ever heard of anybody doing with the Internet occurred last night. I happen to be on the Board of Directors of this condominium, and a proposal had been made from one of the committees about planting some ivy somewhere, some kind of plant cover—I didn’t even understand the proposal—and one of my colleague Board members had looked on the Web to find out about plant covers, and in the process of doing that, she said that she found out from the Web that rats love ivy plant covers! She printed out a whole lot of stuff—which I didn’t look at; she gave it to the people who were doing it—a lengthy discussion about rats and ivy and plant covers, all stemming from having looked on the Internet to find out about ivy plant covering. Amazing! Just amazing.

I’ve reached the conclusion that there isn’t anything you can’t find on the Web if you look hard enough. Now, it may not be correct. I’ve seen some biographies of me on the Web that are pitiful. I mean, they’re so bad that it’s almost sickening. Obviously term papers where the people did things badly.

Abbate:

Has it ever been an issue for you that you had family responsibilities that made it difficult to balance work?

Sammet:

I chose to pursue a career rather than a family, so that was my choice.

Abbate:

Did you have people, either men or women, that you considered role models or mentors, who helped you, either in school or in your career?

Sammet:

Not really. I guess Grace Hopper comes as close as anyone to being a role model, not a mentor. But I don’t think I ever thought of her that way until later, when I thought, “Oh! She was a good role model.” But I didn’t think of that till many years later.

Abbate:

Well, I don’t know if people necessarily do think of it at the time.

Sammet:

No. And, because you may ask, I don’t know whether—I haven’t consciously been in a situation where I was a mentor. Maybe one or two people; I can’t remember. But I have on occasion, just out of the blue, gotten some very strange but very nice letters that said, “Gee, I wrote you twenty years ago, and you sent me back a very nice answer and I’ve always been very grateful.” You know, that shows up once in a great while. That’s always kind of nice.

Abbate:

Interesting. Well, I just have one more question. Do you have any advice for young women, or young people in general, who might be thinking of going into computing as a career?

Sammet:

Learn to write well and talk well.

Abbate:

Hmm!

Sammet:

Are you surprised at that?

Abbate:

A little.

Sammet:

Some years ago—probably about ten or fifteen years ago—I was asked to go speak to some young high school students near Argonne, Illinois. Argonne Laboratory, near Chicago, was organizing something for high school girls—not girls and boys, but just girls—in the Midwest. They were bringing in career women from all different areas of science—chemistry, or physics, or computing—and I was brought in for that. It was sort of a career kind of thing; you know, “What kinds of careers are there?” and “What do you need to do?” and “What kind of problems are there?” I had a number of charts, and I was describing the kinds of careers that can exist: hardware, software, and this, that, and the other thing; and part of my chart was that you need good skills in writing and in speaking. And one of these girls got up in the question period and said, “I want to write programs. That’s really all I want to do. I think I’ll be good at it; I’m good at it now. Why do I have to worry about having good writing skills and good speaking skills?” So I launched into a long speech, which I will encapsulate as briefly as I possibly can, which said, in essence, “If you’re going to present the results of your work, whether it’s internally to your management or externally to a professional group, you have to be able to speak well, and coherently, and so forth. Secondly, when you get to a point where you want to propose certain things to do, that you want to do yourself or that you want other people to do, you may have to write a description of them and present that to various people. You may also have to orally present your proposals to people. There’s no getting around that, unless you want to stay as a journeyman programmer, writing code all of your life.” And I said, “My experience, in observing people—not just myself—has been that people can be very happy programming for a number of years, but after a while, they want to do something a little bit broader. They want to keep their hands in it, but they want to do broader. And that’s when you need the skills for speaking and writing.” And I said, “By the way, with regard to writing, you are also going to have to write project reports, even as a programmer, and you are going to have to document your program. So you need to be able to write competently. I’m not saying you have to be Ernest Hemingway, but you do need to be able to write competently; and you need to be able to speak to an audience and make yourself understood in a reasonable and coherent fashion.” And she said, “Oh!” and sat down. She said “Now I understand,” and she sat down. So I satisfied her, I guess.

I really do think that some people think that what they want to do is go in a corner and write code, and that just isn’t going to last the rest of their lives. It just isn’t. I don’t think it’s going to be enough. There may be a few people who will come out of college at 20 or 22, start writing code, and retire at 65, having written code all their lives. I doubt there are very many of those people, in this day and age. They will want to stay with the programming, with the computer field, but they will want to broaden their horizons beyond just writing code—however much fun that is.

Abbate:

That’s interesting. Well, thank you very much for talking with me at such length.

[END OF RECORDING]