FOSDEM is the biggest free and non-commercial event organized by and for the community. Its goal is to provide Free and Open Source developers a place to meet. No registration necessary.


Interview: Robert Dewar

Robert Dewar will give a talk about "Free Software: A viable model for Commercial Success" at FOSDEM 2012.

Could you briefly introduce yourself?

I am Robert Dewar. I have been involved in programming language design and implementation since the late 60's (working on Algol-68, SNOBOL-4 (SPITBOL), COBOL, and since the early 80's on Ada. I was a professor of computer science first at Illinois Institute of Technology from 1968 to 1975, and then at the Courant Institute at New York University from 1976 to 2005.

In 1994, with Ed Schonberg and Richard Kenner I founded AdaCore, of which I am now President and CEO. This is a 100% Free Software company devoted to building tools for reliable software (in particular safety-critical and security-critial software) based on the use of the Ada language and related technologies. Our implementation of Ada, GNAT, is part of the GCC system.

What will your talk be about, exactly?

AdaCore is a successful company totally based on Free Software (all the software we sell is licensed under the GPL, or other free software licenses). We are a successful multi-million dollar company with over 60 employees (together with our French sister company in Paris). My talk will be about the business model we have adopted, and how it has been successful for us and our customers. And also I will discuss how we interact with the Free Software community in general, and the gcc project in particular in a way that is mutually beneficial.

What do you hope to accomplish by giving this talk? What do you expect?

Show people that there is no conflict between the use of Free Software licenses and building a successful commercial enterprise that can sustain itself over a long period. I believe this model is applicable in many other circumstances.

With the GNAT Academic Program (GAP), AdaCore supports the study of Ada in universities. What are some of the most interesting projects that came out of this program?

Fun examples can be found at our website including robotics projects using LEGO Mindstorm, a real-time model railway system, a Mini Submersible, and a project to orbit a university designed and built satellite around the moon!

Open source plays an essential role in AdaCore's business model, which is not a trivial achievement. How do you explain this successful marriage between open source and business in your company?

That's what the talk is all about, often people think there is a real conflict here, and ask us how we can survive. The message I will be giving is that this is actually a very comfortable coexistence both for AdaCore and for its customers. The business model we have developed over a 15-year period has been successful for us and our customers.

Ada is not as well-known as other programming languages, what are some of its unique selling points?

It's particularly suited to very large scale systems with very high reliability requirements. Some examples are avionics systems (the Boeing 787 uses Ada in its flight control system), and air traffic control systems (the land based components of the new UK ATC system are entirely in Ada).

The SPARK language is a strict subset of Ada augmented with a system of contracts. What's so special about this?

SPARK is aimed at "correctness by construction", making use of formal proof techniques, which are guided by the contracts. The idea is to write software that is correct from the start, rather than writing incorrect code and then trying to debug it into behaving well. An example is in fact the ATC system in the UK that I mentioned above, that is programmed in SPARK, and is formally proved to be free of run-time errors such as what would be called buffer overflows in C.

What are the biggest compilation challenges presented by the Ada programming language and what impact did these have on the GCC backend components?

We did need to modify the back end of GCC, enhancing it with some new concepts that represent stuff that is in Ada, but not in C. For example, variant records, which resulted in the notion of self describing types. Basically though, the back end was well suited to the addition of Ada. One thing that definitely did occur was that Ada placed a general heavy stress on the back end, which showed up language independent problems that needed correcting. In addition Ada uses extensively features that have always been in the GCC back end but not heavily used, for example nested subprograms, and dynamically dimensioned arrays.

GNAT is licensed either under the GPL or the GMGPL (GNAT Modified General Public License) license. The latter adds the freedom to produce proprietary programs using the compiler and run-time library. What's the reason for this special-purpose license?

Well this is copied from the approach with C (and its run time library). This allows GNAT to be used to generate proprietary programs. The FSF has always been more pragmatic than some might suppose. Richard Stallman would prefer to see everyone developing Free Software, but if they are going to generate proprietary software, he would prefer that they do it using Free Software tools. For most of our customers, this ability is crucial. It will be a while before we see software like the embedded avionics of military and commerical aircraft being open sourced, and indeed some of our customers work on classified projects so we don't even know what they are working on specifically.

Have you enjoyed previous FOSDEM editions?

Definitely, and we hope this will be another successful year, and that people will find the talk I will give a positive message that the Free Software approach is viable in the commercial world, and encourage others to follow the same path.

Creative Commons License
This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.