N/C WORLD interviews DOUG ROSS
(Thirty years ago -- March 1970 issue of N/C World magazine)
Mention the name Douglas T. Ross of MIT to any advanced builder or user of N/C equipment and no further introduction is necessary. Mr. Ross was in on N/C programming from the beginning as a member and leader of the original development team at Massachusetts Institute of Technology. He has now formed his own company to continue this work on a broader scale.
N/C WORLD: Would you please fill us in on some of your background in becoming involved with numerical control and computerization?
ROSS: I came to MIT in 1951 as a teaching assistant in the mathematics department, then went to work at the Servomechanism Laboratory in the summer of 1952. At that time I entered the computer field and taught myself programming on the Whirlwind computer and decided to stay there rather than return to teaching. So, I obtained a masters degree without course specification in 1954 and finished all the work for a Ph.D. in pure mathematics, but then didnt complete the program in terms of thesis and exams because when that came up in my schedule we had the APT programming effort going on all over the country with some fifty or so programmers involved as well as another Air Force sponsored project for which I was responsible. I also had three groups of programmers in different parts of the country working on it. So I never went back into the pure mathematics to write the thesis.
At MIT in the early Fifties we started many of the techniques still used in APT programming and computer-aided design and language processing. Their roots go back to airborne fire control systems involving three-dimensional geometry -- very intricate mixtures of logic and arithmetic calculations. In 1956 the continuation of Air Force sponsored work in numerical control was turned over to my group. I had the Computer Applications Group in the Electronic Systems Laboratory, (we changed the name from Servomechanisms Laboratory) and we started out on this programming assignment with nobody who had done anything in numerical control at all, because the previous group had left the Laboratory to form a private company.
N/C WORLD: Just what was your working relationship with the Air Force and aircraft companies in the development of the APT programming language?
ROSS: Our Air Force monitor, Bill Webster, who played a very significant role throughout all of this development by guiding through Air Force channels the long years of continuing support, suggested that the numerical control subcommittee of the Aircraft Industries Association would be able to tell us what was needed in the way of programming languages for machine tools. So I went to a meeting in February of 1957 and gave them a progress report on what wed done, making computer displays of three-dimensional cutter paths and so forth, and I described the general approach we were taking and this seemed very interesting to them.
At this time the Air Force was supplying a lot of machine tools that were supposed to be delivered the next summer and the company peopIe said "Where are we going to get the systems to run them?" So I proposed that they send six or eight people to MIT to work with my beginning group and I thought we could make them a beginning system to start the ball rolling.
Well, it turned out that they couldnt afford at that point to send people to MIT so instead we broke the job up and had nineteen aircraft companies working on different pieces of the system all over the country that summer. Later it grew into about twenty-six companies by the time we finished it in February, 1959, and turned APT over to the public domain. Then it was a year or so later that the Aerospace Industries Association (they also changed their name) set up the APT Long-Range Program, and from there on, APT has had quite a history.
N/C WORLD: And this led you directly into the automated engineering design concept?
ROSS: The last thing that we actually contributed to the APT effort was to supply the arithmetic element program, the analysis and program that does all the complicated cutter path calculations. During that same year we were making various studies at the beginnings of what we were calling computer-aided design. We did some experiments on language processing and data structuring. We had developed new methods of representing problems in computers for this cutter calculation for APT and we saw in that a generalized way of representing all problems so we could have a uniform treatment for calculating or data processing or whatever, all in a single framework generalizing the idea of list structure -- representing relationships between things by pointers.
So we determined that that would be the starting place and I can remember even in those days talking with the San Diego APT programmers, saying that our first step would be to make a language and a compiler so that we could do directly the things that we were hand programming.
N/C WORLD: How can you tie this in to direct applications by the user?
ROSS: One important thing we had learned from APT was that when you develop advanced technology, if you dont help users to understand it as you go, you can very quickly come up with exotic techniques which nobody knows how to use or apply. So we started a cooperative program, again inviting companies to send people to MIT to work with us. The name of our system was AED (Automated Engineering Design) and we had quite a program.
We never did a great deal in the way of applications of computer-aided design. We were concentrating on the tools and techniques necessary to build design systems, because we felt that we should not specialize in any particular type of problem. We felt that the people who had the problems in industry would understand their own problems much better than we did. But, we could supply tools that they could then learn to apply to those problems, so we were always working on aspects that were intended to support real production and design problems. We did very little direct application, but others did.
N/C WORLD: So you are actually talking about a system for making systems?
Just as the machine tool
industry has high level techniques and specialized tools, for making
production tools that then produce products, our programming systems
and languages were intended to produce the programming systems and
languages that then would be used to make products. So it is one step
removed from a direct application system. But we did use these same
techniques for making the systems themselves.
So we talk about "a system of systems for making systems."
N/C WORLD: But with so many languages for so many different types of machines, isnt this concept rather a difficult one?
ROSS: We designed all this work so it would be machine-independent to a very high degree, and in fact all of the systems are written almost entirely in their own language so that this makes possible to process called "boot-strapping," picking yourselves up by your bootstraps, in which you have a compiler for the AED language working on one computer, and you change its output part, so it will compile programs for a different computer. Then if you run through that compiler the language statements that describe the compiler itself, you come out with a version that will now run on the other computer. Then you have to add in some hand fixes for the parts that dont carry across properly -- but they are very few in number. So this is the process whereby we get these tools for making software systems.
N/C WORLD: Just what prompted you and your cohorts to leave MIT and develop a private cooperation for doing this type of work?
ROSS: It became clear in late 1968 that we were unable to supply all the things that the cooperating companies needed in order to make use of these techniques. We had people come and work with us but when they went home, for various reasons the actual employment of these techniques was very spotty -- some places went very well, some not at all. The reason was that this advanced software technology requires a good deal of sophistication for the initial application and from the MIT environment we were unable to supply direct handholding assistance.
At the time we were finishing up the third version of the AED systems, called. Public AED, and were scheduled to turn it over to industry at the last public meeting of the MIT project on July 15. We decided that that would be the opportune time for us to set up a private company to give the help that couldnt be given from the academic environment.
Also, it was clear at that point that the government funding would not be as substantial as it had been in the past, although it was not disappearing entirely, by any means, and we already had obtained additional new funding from NASA. But the whole combination of things came together to say this would be an appropriate point -- that if we didnt take this step it was likely that all these ten years of effort would essentially end up being an academic exercise. We would only leave the users still farther behind. So we took the plunge.
N/C WORLD: Just who should be the subscribers to what you have to offer? Would they be primarily users of machine tools under some sort of control? Would they be builders of that type of equipment? Or people with product design problems? Or all three?
ROSS: Actually all three and then quite a few more because the neat thing about this whole software technology capability is that it isnt specialized to any particular class of problems -- which I think is particularly appropriate for todays N/C world in the sense that more and more its becoming recognized that its the full control of the business enterprise in which numerical control plays its role. We plan to have a variety of focus points -- specific application areas -- so a sponsoring subscriber can find several areas of interest in addition to the general public AED itself.
N/C WORLD: But SofTech doesnt really have any proprietary information or products as far as automated engineering design and so forth is concerned.
ROSS: Thats right -- not at this time. SofTech has only one proprietary product at this point -- an on-line debugging system for Fortran programs -- not related to AED. We sort of feel like custodians of the AED work, however. Almost six million dollars went to support AED work from the Air Force and company participation. The compilers, systems and so forth would cost between two and three million dollars of research and development to duplicate now -- and its all there for people to use.
So we feel very responsible to show people how to use AED and keep it going. We know how to use it better than anybody else at this point, but have no proprietary hold on it.
N/C WORLD: So then, just what are the objectives of your company and its programs?
ROSS: The objective is to bring a man and a machine and a problem together to get the solution the man wants, and thats all. Its applicable to almost every situation where this appears -- business applications, inventories and job scheduling, as well as numerical control, graphics and mathematical scientific computations.
The actual techniques that are available work equally well for banks or insurance companies. The tricky thing is to have the given company understand what it wants to do well enough to see how it can be done in a systematic way that allows a language for that class of job to be designed.
In other words, if you think of a single problem, then its better to just program it immediately but if you are going to have a large family of similar problems -- such as a family of parts to be designed or a family of machines with related actions to be controlled, or a family of scheduling jobs all of which are similar but have their own differences -- then the way to obtain maximum benefit from using the computer in such environments is to capture the essence of that class of problems in a system, and have all of the difficult parts automatically available in the form of little building blocks inside the computer. Then somebody who doesnt understand how to program computers or may not understand the needed mathematics can talk just in terms of the problem itself, in a user-oriented language. The translation of that language is done by another computer program, which automatically assembles the proper little steps of action to solve the particular problem that he had in mind.
That is all that APT does, for example, with numerical control. The parts that are produced are made up of straight lines or circles or motions of one sort or another and by different descriptive techniques the part programmer is able to put these little actions together to produce arbitrarily complex big action. So its to get the same effect for many application areas that these techniques are particularly appropriate.
N/C WORLD: What are your actual marketing approaches now?
ROSS: Marketing is difficult if we try to talk about the full use of the technology that lies behind everything. So we actually break the activities of the company into four related areas which come together to make the over-all abilities make sense. One is just plain high level consulting. We have a great deal of experience in a wide range of industrial, engineering and scientific application areas in time sharing, graphics, and so forth. We can also apply many of our system concepts merely to assist people in understanding their problems better.
Coupled with consulting is our education program. SofTech will present a series of formal courses, with both lectures and workshops here and in various places around the country, as well as on site in plants. We will instruct programmers in how to use these techniques, and instruct managers, how to understand what they can do. We also will design special education for particular systems as we go along.
But the primary marketing thrust at the beginning is concerned with undertaking to build, under contract, systems to do particular jobs, using these techniques. AED needs more visible applications now.
N/C WORLD: So you are actually working on a combination of systems which allow you to define a new language along with the building blocks that prQvide most of the makeup of translator for that language.
ROSS: Right, and that is what we started to do ten years ago, because we said you couldnt have one computer-aided design system to serve all problems. It was absolutely insane, and you wouldnt want it if you could have it. What you really need are many, many very highly specialized languages, each one attuned to its own application area so that people that dont know a thing about the details of the job they are doing nor much less how to do it with a computer, can still be able to solve problems. And thats where the user-oriented language comes in. And of course each of these classes of users is different. So youre faced then with the problem of producing hundreds of specialized languages, and ordinary programming techniques would involve tremendous expenditures of manpower and money for any decent programming language system built by ordinary techniques. You spend many hundreds of thousands of dollars to write such a system, and furthermore, the skill to write the systems just isnt around.
So we must automate the production of languages, the producing of language processors -- otherwise we cant tie these people together with their application. So, this is what software technology is able to do. Most of the people who write software systems still produce them by hand.
That is why they are expensive and late and costly, and dont work well, and they are rigid, hard to change.
You were asking about what kinds of companies would be interested in this new technology. I hope its clear that it is anybody who has a problem. With this high level technology, youre able to take such a new application area and almost overnight come out with the first system you can turn over to users to experiment with. You put very little expenditure of time or money into it because you have assembled most of it out of ready-to-go software component building blocks. After user acceptance the same technology will make a refined production version.
N/C WORLD: Where is all this going to lead us? Can we ever standardize on languages?
ROSS: Some people worry about standardization -- when they think about all these different languages. They say "Sure, each individual group is happy because they have their own specialized language. But how are they ever going to talk to each other? We dont need the tower of Babel. Wouldnt it be nice if we could have one language?" No, it wouldnt, and its not necessary either.
The key is that is all these languages have been produced in systematic ways, based upon a workable, technological basis, they will be able to be translated from one into the other where appropriate. They arent really foreign. Theyve all grown from a similar family of languages. So that where its important for different groups to communicate, they will be able to do so, but when they are working within their own area they can work most efficiently there too.
N/C WORLD: Are the potential users of these techniques really taking aovan tage of them to any degree at all right now?
ROSS: Not to a great degree at all right now. That is the major reason why we decided not to continue further research advancements at MIT.
N/C WORLD: What bit of advice would you give to the readers of N/C WORLD with respect to all that we have discussed?
ROSS: I think I would point tb the pervading impact of numerical control already. Look at its history -- how by its nature it spreads far beyond the actual milling, reaming, boring action of the tool and spreads throughout the individual company -- the whole business, in fact the whole economic structure. A tremendous range of things are influenced by this coupling of automatic control with the minds of men.
I think your readers should realize that the kind of performance that now is taken almost as commonplace for immediate applications of numerical control is now able to be extended to all aspects of doing jobs, whether they are paper work jobs or jobs requiring muscle.
The key thing is to have each kind of job best defined both for itself internally and for its relationship to other jobs. A business is an intricate complex of many kinds of jobs that go together to make up a total production machine including the people as well as the mechanisms.
These relationships can be captured and made more reliable, getting the human errors out, and getting the human understanding into the system. We can only expect that there will be greater benefits for all of mankind coming out of these developments.
(This got cropped off from the bottom.)
from the March '70 isssue -- Copyright 1970 by N/C COMMUNICATIONS, Inc. the world of manufacturing with numerically controlled machines
(I don't yet know how to have the text run around this)