Ada helps churn out less-buggy code
Ada helps churn out less-buggy code
S. Tucker Taft
S. Tucker Taft, who led the Ada 9X language design team from 1990 to 1995, is now president of the Ada Resource Association, organized last year to maintain the language standard after the Defense Department closed the Ada Joint Program Office.
Since 1980, Taft has worked for AverStar Inc., formerly Intermetrics Inc., of Burlington, Mass. He is technical director of the distributed information technology solutions division.
At AverStar, he has headed development of three new compilers based on Ada95. The first generates Java byte codes, the second creates optimized ANSI C, and the third generates code for the Sharc digital signal processor from Analog Devices Inc. of Norwood, Mass.
In the 1980s, Taft helped design the Ada Integrated Environment for the Air Force. Taft, the great-grandson of President William Howard Taft, graduated summa cum laude from Harvard College in 1975 with a bachelor's degree in chemistry.
GCN senior editor Florence Olsen interviewed Taft by telephone from his Massachusetts office.
GCN:'After its long and checkered history, why do you think Ada is still a useful language?
TAFT: Ada is good at detecting errors in programs early in the programming lifecycle. You submit your code to an Ada compiler, and it will inform you right away that you made mistakes that a lot of languages would let go right by. For some people, that's the bottom line. Errors cost money. The sooner you catch them, the less they cost.
If the compiler misses an error, Ada has another line of checking when you run the program. People who use Ada generally find that if the program makes it through the compile-time and run-time checks, it's remarkably close to doing what it should. That gives them a sense of productivity and pride in the quality of their work.
GCN:'Does the Defense Department still share your positive opinion of Ada?
TAFT: I hope so, but I wouldn't say actions recently have backed that up. The problem is, DOD is a huge organization that has a three-year total turnover at certain levels. Someone who had a good experience with Ada may have moved on, so you end up having to re-educate over and over.
Some big military programs committed to Ada are just getting to the coding stage, so the number of compilers being sold into military programs is on the rise.
Joint Strike Fighter, various tank programs, the Apache helicopter'probably 50 programs, maybe more'are still actively developing in Ada and spending millions of dollars.
GCN:'What is notable about the software companies in the Ada tools market?
TAFT: They have all had a long-term commitment to Ada and still believe in it. Rational Software Corp. of Cupertino, Calif., is the biggest and has been one of the leaders in Ada since buying Verdix Corp.
GCN:'What about the current Ada job market?
TAFT: The buzz seems to be that people are looking for Ada programmers, particularly for work on DOD systems that are ramping up. The thing I tell people is, if you're looking for an Ada programmer, just look for a really good programmer. The Ada compiler helps you learn the ins and outs of the language because it doesn't let you give it any old junk. It's picky and instructive in terms of guiding you in proper use.
Ada has many features that push you toward thinking in a modular way using abstractions. That's true with a number of other object-oriented languages as well. People who are familiar with programming in other languages find that with a good Ada compiler, they can become productive in a couple of weeks and start churning out programs with fewer bugs than they could in their old language.
GCN:'Do you expect Ada95 will remain the only standardized object-oriented programming language for the foreseeable future?
TAFT: No, because I believe C++ already joined the crowd about six months ago. There is no official validation suite for C++. Within the Ada community, there is more interest in implementing the standard exactly as it is written.
GCN:'What is the federal government's best Ada success story?
TAFT: These days we're focused mostly on commercial success stories. But in terms of high-profile uses, the F-22 and Apache helicopter are programs people have heard of.
All the military jets in any of their recent versions are programmed mostly in Ada.
When you get into an embarrassing situation, it's usually because the system got too complex for the engineering tools being used. Ada doesn't seem to have that kind of limitation.
If you look at why Microsoft Windows 2000 is a year and a half late, a lot of it has to do with bugs. Whether it would not have had the bugs had Microsoft Corp. written the millions of lines of code in Ada, I don't know. But it's clear Microsoft is getting beyond the limit of what can be managed in its engineering environment.
GCN:'What is the average percentage of code reuse in any given Ada project?
TAFT: NASA did a systematic assessment and measured between 60 percent and 80 percent code reuse from one satellite system to the next. It was comparing reuse against past use of Fortran, which was NASA's primary language for satellite systems. With Fortran, code reuse was about 25 percent to 30 percent
I think NASA concluded that Ada is not magic, but that the notion of a package with a clearly defined interface is what you need for reuse. As NASA has taken what it learned from Ada and applied it back to Fortran or C, it's been able to push up the reuse.
GCN:'Does Ada contribute to organizational productivity as opposed to programmer productivity?
TAFT: When you start bringing the pieces together, they fit because of all the checking that Ada does. Other languages that do less interface checking don't always work together when you try to put them together.
GCN:'Are you bullish about widespread adoption of component software?
TAFT: It's a little hard to say anything about what people will do in the future in the programming world where there is a tremendous amount of faddism or hype. You have bubbles that grow and then pop dramatically or just lose air.
C++ was it for five or 10 years, and now it's Java. Ada has been a nice solid performer through the rises and falls. We've also made a good effort to integrate Ada with these technologies. There are Ada compilers that connect into Java Virtual Machines and compilers that connect into C and C++ compilers and so forth. Nobody's going to build a whole system in any one language any more.
One of the biggest problems is logistical. How do you sell the components or make certain they are what you want to buy? But I think that will be worked out. There is no reason the components themselves cannot be written in Ada, and I know people who have written components in Ada that work in the Microsoft Component Object Model and the Common Object Request Broker Architecture.
GCN:'Can component assembly extend the functionality of existing systems, or does it work only with new systems developed from scratch?
TAFT: People have figured out ways to make a legacy system look like a modern component through the art of wrapping. They've discovered the idea of a really good interface. You can hide almost anything behind it. Ada is one of the first languages that focused on the interface as a packaged specification.
GCN:'Are there other issues that could kill the component assembly approach to application development?
TAFT: Some issues are legal. For some kinds of applications, particularly those that involve safety, you need to write everything yourself or have very close control. I think we'll end up having some specialized vendors that certify their components in the same way they might certify a brake assembly.'If there ever is a component market where you need component warranties, I think Ada will be a big player.
GCN:'Do you have any preferences among the component architectures?
TAFT: COM is focused on getting the job done one way or another in the usual Microsoft build-it-and-then-figure-out-what-the-design-was fashion. CORBA is design it first, then have multiple vendors implement the specification. I prefer the second model, but I understand the value of the first. They're converging on one another.
GCN:'How new is the idea of instrumenting applications to monitor other applications?
TAFT: In a component-based world where you might have 20 components from 20 different companies, an awful lot of finger-pointing can happen.
One of the things we've been doing at AverStar is developing tools that help identify where problems are.
- Family: Married; two daughters, ages 8 and 12
- Car: Dodge Caravan
- Last book read: Private Screening by Richard North Patterson
- Sports and leisure activities: Skiing and tennis