Mary Van Deusen:
The Ada Language Contest
Coming to work at Intermetrics was the most incredible luck. My background had been as the baby of the systems development group at University of Chicago and then, when we moved as a group to National Cash Register in Los Angeles, as a maintainer of NCR's Rescue/Restart system. Systems programming struck me much as LA did - lonely. It was machine-focused, rather than people-focused.
I'd gotten some interest in languages when I talked my way out of taking an exam for my master's in education, and talked my way into writing a thesis on John Backus' Reduction Languages, or Red Languages as they were called, as a vehicle for teaching children about computers. It had made such sense. From experience, I knew that variables were a source of confusion for the very young ones, so surely a language WITHOUT variables must be more kid-friendly. And it was, according to the curriculum I developed, until somewhere around 6th grade. After that, your students would do better with a solid master's degree behind them. Scratch one idea.
When my husband and I moved to the Boston area for him to get his PhD at MIT, I began working in computer education - writing scripts about computer topics and having an art department to help turn my words into animated images. But the company moved to Kansas City and I, again, turned my mind to what I wanted to work in next. And, luckily for me, Jim Miller of Intermetrics, my boss, had been told he had to get someone else to write his CS-4 reference manual. It turns out that I wasn't the first to be hired for the job of writing his book on the Navy language. I was just the first that actually got to do it.
At that time, there must have been about 40 people in the company. It was still small enough to keep Swiss Miss chocolate in the kitchen area for chocolate starved employees. The offices were incredibly elegant, as the company president's wife ran a decorating company. Desks were butcher block expanses of oak on brightly colored filing cabinets, and the chairs were huge squares of black leather. Life was, indeed, good.
After finishing the book, I waited for Jim to tell me what he thought. And then waited some more. Finally I went and asked him. He stared at me with some confusion and said he didn't think anything about the book. Every time he wanted to know something, he'd go to the index and find it. It took me a bit of reflection to realize that I'd just learned something very deep. If you do your work right, then you disappear from view and your work stands on its own. It's a lesson I've taken to heart, and a lesson for which I still thank Jim.
After CS-4, I worked with Jim Pepe on the SPL/I Navy language for signal processing. At first he'd make a language change and tell me where I'd have to change my book. And then I started seeing the pattern, and I could take his language change and realize all the second-level ripples that the change would cause throughout the document. It was a long time before I realized that it wasn't just that I knew my book cold. My book WAS the language, and I was finally learning something about language design. The icing on my language cake was when the Navy put my book into their Request For Proposals as the quality of work required for future products. That was pure joy.
Intermetrics was an incredible place to work. On the side of the company where I was, most people had their PhDs and were passionate about their interest in the purity and beauty of the languages they built. It was a company of nice people. They were always ready to explain, and to share, and to laugh. It wasn't a 9 to 5 job, and I would never have wanted it to be. If you would have invented a perfect place to work, you would have had to invent Intermetrics all over again.
In such a heady environment, it was inevitable that I'd feel the need to go back to school, and I enrolled
in the evening program at B.U. My first teacher of languages was Lyn Bates, a BBN researcher with her
specialty in voice recognition, and her PhD out of Harvard. Which means that BU gave me both a master's
degree and a very dear friend.
BU has one more thing I have to thank it for - the acceptance of my career in computers. From the time
I built a telescope at the age of 16, I'd always assumed I'd go into astronomy. But U of Chicago didn't let you take
any astronomy classes as an undergraduate, and the pull of physics just wasn't there for me. I eventually
changed my major just before graduating into the history of art, near eastern ancient. But the dream
of the lost pathway in astronomy was always there. In the PhD program, I had a chance to take some
electives, and I instantly chose an astronomy course. It took only a few classes to realize that my
heart had, indeed, healed, and that it was computer languages I really wanted to study. But without
that opportunity, I might have never known.
Languages pulled me, where systems never had. To work in
languages you have to put yourself in the place of a user and continuously ask yourself what would
make their use of a language as stress-free and natural as possible. Instead of the machine orientation
of the systems world, I found myself centered in a world filled with the beauty of finite state diagrams,
as well as the joy of connecting myself to everyone who would use the language or read the book.
Computers had finally become my home.
When the contest began for creating the common language for the Department of Defense,
the entry from Intermetrics was headed by Dr. Ben Brosgol, who acted as both
project leader and main designer. Ben was trained in the Harvard school of
language design, and was a gentle soul of deep understanding and knowledge.
Under Ben's leadership, Intermetrics was chosen as one of four companies to
build a language in what would be an anonymous competition - thus the color coding
of the languages submissions.
We were the RED team, and Ben called the language REDL.
Of the other three languages, the BLUE language team was headed by
John Goodenough of SofTech, our main language competitor in the Boston
area, though we tended to work on Navy and NASA languages, and SofTech
was fairly entrenched with the Air Force. The other two languages, Green and Yellow,
were designed by Honeywell Bull, the French affiliate of the US Honeywell company, and
Stanford Research Institute, respectively.
When the design phase was over, we went into review mode, anxiously
examining our competitors. What became clear was that there were two
loci around which the designs centered. Blue and Red were designed by
people with experience in the military contracting field. The designs
were both conservative. They were meant to be solid and reliable languages, but
they didn't chew up the language field scenery. Weighing one
against the other, Ben's document won over John's, in part, because John
was such an honorable soul, with his own deep knowledge of language design, that he
felt he had to tell you EVERYTHING. And he did. There were pages
devoted to the meaning of BOOLEAN. And it was true that he described the
actual world down there in the implementation nitty gritty. The problem
was that the reviewers weren't as skilled at reading what John was skilled
at writing. Certainly not in the very short time they had available for their review.
John's books were immense, and the review time absurdly short.
Red won over Blue because people could understand Ben's book better with a
quick and dirty read.
Yellow and Green were something else altogether.
They seemed the products of academics rather than of engineers, though the head of the Green
team, Jean Icbiah, had a strong background in industry and practical experience with his
design of the LIS language.
Both teams invested heavily in very famous consultants, and both languages were ambitious in their goals.
Practical implementation was not a first consideration, but the languages promised excitement.
Green had one other thing that none of the other three had - a fine-tuned political sense. It was said that Green's first design decision was the length of the book they would create. Where Blue had spent pages on the real meaning of Boolean, Green said there's true and false. Period. Green demanded that you trust the designers when they said what the language could do, even if the details weren't there to prove it.
In the end, the reviewers chose to trust the Green team. But they needed a second place finisher, and a workmanlike language was a good safety backup. So they chose one from column 1, Green, and one from column 2, Red. But it was always assumed that Green was going to win. They just needed enough time to flesh out what everyone assumed Green knew about their language, and just hadn't told them yet. As well as to build a translator that would prove that all their words weren't just smoke and mirrors.
During the first contest phase, Ben had worked his tail off. Everything else had had to take a back seat,
including his pregnant wife. Coming in second meant that the pressure would start up again
after too short a break. Besides, having everyone assuming we were going to lose took a
lot of joy out of the coming work. So, when one of our other main language designers returned
from a sabbatical teaching in Germany, Ben let John Nestor take on the heavy lifting so that Ben
could spend more time lifting a beautiful new baby son. Ben assumed that John would do what
Ben would have done, which was to fine tune the first phase design to fill in what holes had been
noticed in the design. Ben should have known better.
To John, it was still a contest, and one that he had no desire to lose.
He recognized the excitement of the Green language, and realized he wouldn't have a
chance in this phase unless he could make Red more exciting. His answer?
To throw out REDL and start again from scratch. Recognizing that this approach could
give us a better chance at winning in the end, the rest of the design team prepared to dig
in for another frantic race.
We spent that second phase running flat out. But at least we had a goal that we could see in
front of us - the Green language, as of the end of the first phase.
The intention with the new Red was to build a sound and elegant skeleton on which the new
language would sit. In the early days, John avoided syntactic sugar altogether.
This approach might have been more efficient, but it did make for a bit of stomach-churning.
It was unfortunate that we were required to give an intermediate presentation in the middle of
the redesign phase, as the language that the reviewers got to see
was much more complex at that point than it would be when it was complete at the end of
the redesign phase.
What we didn't really understand was how convinced the reviewers were that the contest was essentially over. While we were working insane hours, there were reviewers who were trying to have out the first book on Green, the first tutorial on Green, the first anything on Green. Seeing Red at this intermediate time just convinced them that they were making smart, strategic decisions to take the publishing world, etc. by storm.
Even the team worried about whether the language was going to come together in time. And think about the poor team under Mark Davis' skilled leadership. Not only did they have to build a translator for a language still being defined, they had to build one so flexibly that they could pivot on a toe tip if John felt he had to make some fundamental change for the sake of the integrity of the whole. That team never got the kudos they should have for what was a brilliant piece of work.
Even in those last weeks, the team was holding its breath. But John knew he had a deadline, and he was now building up the sugar decorations on a foundation he could completely trust. On the day that it was due, John was done. Red was ready to be presented for review.
The presentation took place at a church in Arlington VA. Somehow it seemed appropriate, because there had been a lot of tired prayers sent up. We were excited. People came up and told us they had gotten on the plane to the meeting convinced that Green was the better language. Then, for the first time, they read Red on the plane and became convinced that Red was the better language. We were making converts! But, oh, too slowly.
There were two events that day in Virginia that shook us to the core.
One was about our translator; the other,
a voting issue.
Near the end, we were hearing rumors of problems with the Green translator that turned
into the news that Green was not going to have their translator ready on the day it was due.
That should have given Red the advantage, right? Wrong.
According to the contract for Phase Two, the Test Translator was not
to be considered as a factor in the choice of the winning language
design. At Intermetrics we always wondered if the DoD wrote the
contract in this way with the thought of helping Red but, if so, it had
precisely the opposite effect.
Because a discussion of the translator was not to be part of the evaluation, we couldn't
talk about it during the presentation of the two languages. So, when a
member of the Green team get up to tell the congregation that they should choose Green
because Green was implementable and Red wasn't, we were gritting our teeth. After all,
Red had a translator there in Virginia,
but we weren't supposed to defend ourselves by saying that our translator worked!
The nail in our coffin was when the Air Force sent a low-ranking young man to attend the voting meeting. He explained that he didn't really know anything about the languages, but he was instructed to vote for Green, to not discuss the matter, and to not change his vote. Did I mention that it was required that the winning language be chosen unanimously? I leave it to you to imagine our frustration. The deadline had not really been the date to which we had worked. Our deadline had passed before the second phase even started.
We did finally learn something about politics. In order to put ourselves in line to continue getting funding for what was then being called DoD-1, we had to let Red disappear quickly, and completely. And so it did.
And that's why, on this 30th anniversary of the delivery of Red, it seems right that history
should look back at a runner-up language designed by John Nestor and
Ben Brosgol and
Mark Davis and
David Levine and
Michael Tighe, Peter Belmont, Toby Boyd, and many more of us, both in and out of Intermetrics.
As for the Red Reference Manual, I have a fondness for it since I co-wrote the book with John.
I always took great pleasure in the comments of some reviewers that our book was the better of the two
books, whether or not they preferred our language. I've put the entire reference manual here for your
curiosity and, hopefully, pleasure. You'll find my SIGPLAN article on Types in Red
here, as well.
When I left Intermetrics for Prime Computer, I thought that my association with DoD-1 was over, but Prime let me still attend the Ada Implementors' meeting in New York City. While there, I listened to someone get up and explain just how much money was in the offing and how it was the people in this room who were going to get the contracts because we were the only ones to understand this complex language. I walked out of that room fuming! It seemed fundamentally unfair that the knowledge would stay so close because everyone should have had a right to learn and compete for the coming contracts.
The more I thought about it, the more bothered I became. At the time, I was on the Executive Committee for ACM's language group, SIGPLAN. I'd never been an editor of anything, but I figured I could learn. So I solicited financial support from Prime to create a free newsletter for anyone who asked. Since that gave Prime some exposure it wouldn't have otherwise had in the Ada community, they agreed.
The magazine started off with instant community support. As well as a wonderful Ada program written by Bob Mathis for the first issue on how to take care of his new baby. I gathered together a set of department heads and became quickly close friends with all the express delivery drivers. It was a matter of pride, I fear, that no department ever came in except by last minute overnight express. The UPS driver still honks when he sees me in the yard.
Given that the newsletter was free, we developed a good sized subscription base very quickly. Prime couldn't keep up the burden indefinitely, so I began searching out other companies to cover either the duplication or the mailing cost. Since I'd run a lecture group on programming languages while at Intermetrics, I was well acquainted with begging for corporate support.
Eventually I started to worry about how the newsletter would make it when my interests turned me to something else. We had a publication for SIGPLAN, called SIGPLAN Notices, and I proposed to SIGPLAN that we take on my newsletter as a second publication. I was able to negotiate one free year for my current subscribers, and Ada Implementors' Newsletter became Ada LETTERS.
I expect to be adding some of the early Ada Implementors' Newsletters to this site in the future.
A few years later, I found myself serving on the SIGBOARD, ACM's overview board governing special
interest groups, still on the executive committee of SIGPLAN, and still editing Ada LETTERS.
The meeting I'm remembering was held in Nashville at the Grand Old Opry Hotel.
We were up on the 2nd floor sitting on a very long bench.
On one end were other SIGBOARD people, and on the opposite end of the bench were the Ada people. The middle of the bench I
considered representing the SIGPLAN organization. The goal of this get together was to ask for permission
for Ada to break loose from SIGPLAN and become their own separate technical committee - AdaTEC.
Since I was on all three of these executive committees, I would scoot from one end of the bench to the other as I'd negotiate in my three varying hats. SIGPLAN did let Ada go, and SIGBOARD did approve our spin-off.
In 1985, I was invited by Time/Life to consult on their book on computer languages in their Understanding Computers series. I actually did have something that I wanted to say to them. It all whips back to a time when I noticed an Air Force Colonel at an Ada meeting. It was Bill Whitaker, and I was instantly taken with the man. With all the publicity and excitement of the early Ada days, what I remember most was Bill saying that people would remember Dave Fisher for his metal man requirements, and Jean Ichbiah for the design of Green, but no one would remember him. Bill was the one who had made the whole competition happen. That hit me deeply.
With Time/Life listening, I told them all about Bill and his dreams. When the book came out, I searched out the Ada pages and read them quickly. Then I called Bill Whitaker and read him all the pages that told his story. "Now they'll never forget you," I told him. That still ranks for me as one of the things I'm most pleased with having done.
That isn't probably what most Ada people remember when they think about that section of the book.
The question I inevitably heard was who was it who cried themselves to sleep from exhaustion on
the Red team, a quote from the book. Yes, dear reader, that was me.
But I wouldn't have missed it for the world.