Aad van Wijngaarden, the founding father of informatics in the Netherlands

Adriaan van Wijngaarden was born November 2nd, 1916, in Rotterdam. His brothers were like him, smart and energetic, but Aad was the favorite. He was a precautious young lad and, remember, he had a symbol of his own on the typewriter. For him the way was cleared. He was an eager student of nature and wildlife. He knew every bird in the sky and posessed his own collection of minerals. He would read just anything, had a fascination for language. A true know-it-all, for him knowledge was a formidable, sometimes fearsome, weapon in social interaction. He attended classical grammar school at Erasmiaans Gymnasium and subsequently got the opportunity to enroll at the Technische Hoogeschool in nearby Delft. His father died young, in 1928. Higher education was not a self-evident choice. Yet, Delft, the one technical university in the Netherlands, was literally at the horizon of Rotterdam.

Aad studied mechanical engineering and specialized in naval construction. The design of ship propellers was his graduation topic. Exam regulations, tuned to a Polytechnics-style tradition, had it that an engineering degree should be based on practical work. By the 1930s, however, engineering studies had grown increasingly theoretical. The bigger challenges were in fields like Applied Mechanics, represented by professor Cornelis Biezeno, and Aero- and Hydrodynamics represented by professor Jan Burgers. Scholars of international acclaim, in Delft they were a subculture, deemed too theoretical to have students of their own. It was this subculture where Aad van Wijngaarden found a home. Ship propellers were a typical Burgers’ topic. He found a fellow student to jointly study the full two volumes of Courant and Hilbert, Methoden der mathematischen Physik. He devoured Biezeno and Grammel's all-encompassing Technische Dynamik (1938).

After graduation in 1939 he was hired as an assistant for research in naval construction. He was responsible for the towing tank at the section of naval engineering. This task, combined with assignments for Biezeno and research for Burgers, sailed him through the war years. His first publications ensued.

Burgers was the man of turbulence, i.e. the theoretical model of turbulence, later known as Burgers’ equation. In the same vein he asked Van Wijngaarden to study the possibility of air bearings, a thing that would rotate so fast that it would float on air. Modeling this might come of use, for example, in the design of an ultracentrifuge.

It required the handling of third-order partial differential equations. It took him years, weeks and weeks of cranking before another instance of the equation would be solved. The work was only alleviated by permission to take his Marchant electric calculator home so that he would not have to go to the laboratory during the freezing cold winter months of 1944-45. When in 1945 a portion of the work had been completed, Burgers advised to write down the result and defend it as a dissertation - leaving Van Wijngaarden in deep frustration: “It simply lacked beauty! That is, I could see how to continue but all of the calculations did not yield the kind of insight that I wanted.”

How could his teacher, his lead in judging good research, ask of him to be content with such an ugly result? Aad van Wijngaarden dropped the whole project and never returned to it.

The delicate situation came to the attention of Biezeno, who did not hesitate to come to Aad's rescue and created the opportunity to complete a doctorate based on his elaborations of the Biezeno and Grammel book.

The novel dissertation Enige toepassingen van Fourierintegralen op elastische problemen brought him the doctorate with highest honors, cum laude, on December 12th of the same year 1945. One of the propositions he chose to defend, read: “It be recommended to offer students in mechanical engineering the opportunity to graduate in theoretical directions.”

It goes without saying that every single one of the reports and certainly the dissertation was delivered in person to his mother with a personal note, signed with his name, spelled “@”.

Burgers on his part certainly did not abandon his pupil. It was upon his suggestion that the Delft Research Fund decided to send in 1946 a dozen of young engineers to the United Kingdom to catch up with the research they might have missed during the wartime and collect as many research reports as they could gather. Indeed, research had in part continued, even in Delft under German occupation. But more importantly a major shift took place in research. As of May of 1945, it abruptly changed language from German to English. With the change of culture new connections had to be built up. Aad van Wijngaarden was the one who was sent out to cover the fields of naval engineering and applied mechanics. He would report to Biezeno, weekly. The report of January 22 conveys a remarkable decision.

Writing from Worsley House Hotel, London, he could not hide his enthusiasm on having met A.C. Stevenson and W.M. Shepherd: “cf my dissertation, page 62, note 1”. Computational methods instantly turned into real living people. Stevenson showed him the ropes of computational culture in England, and he eagerly absorbed the knowledge. They became friends, whom he would later visit together with his wife. Two weeks on, upon meeting Shepherd again in Bristol he noted: “I would prefer to report separately and more coherently on all that I have learned on mechanical aids to computation, like differential analyzers, calculating machines, and punched card systems.” By consequence, February 8, 1946, may well count as the date of birth of computer science for the Netherlands. His letters offered a series of small observations from experience, e.g. that Marchant machines are easier to handle with than Friden machines; that the arrangement of work was different in the UK from what he had seen in Dutch laboratories; that the feared blue booklet MTAC, Mathematical Tables and other Aids to Computation, had recently appeared as a journal.

The final trip report did offer a separate section on the “mathematical institutes”, homes of computing practices – Manchester, Cambridge, Teddington, London.

Leslie Comrie was the most experienced computer of them all. Emerging from the Nautical Almanac Office, his Scientific Computing Service in Londen was a private enterprise since 1935 that could sustain itself, delivering computations to order for public and private customers. Van Wijngaarden was deeply impressed by the variety and the sheer number of machines installed at the service, and its comprehensive library of mathematical tables. He learned every day.

Implicit in maturing in the subculture of Biezeno had been knowing how to compute and, more than that, knowing how to set up and execute computations. One would know how to grasp a phenomenon, describe it, model it, attribute an equation to it in such a way that computation was feasible. Feasible computations would be expressed in computation schemes and well-known schemes of computation were printed as computation forms, “Rechenformulare”. The achievement expected of an experienced scientific computer would be to not only know one's way among the various schemes but also to design novel schemes if needed. Being part of that culture, one could tell the difference between beauty and ugliness. Van Wijngaarden had the experience and was able to recognize the English techniques and technologies of computation. In a timespan of only weeks, he was able to familiarize himself with the decisive extension of his know-how.

What during his studies had only been helpful approaches towards advanced research, now, through this visit, became a an expertise in its own right. German scholars would discuss the expertise in terms of computational procedures, “Rechenverfahren”; Carl Runge of Göttingen considered it the executive power of mathematics, “die Executive der Mathematik”; Alwin Walther in Darmstadt called it “praktische Mathematik;” while Mauro Picone in Rome called it “applicazione del calcolo;” to Friedrich Willers in Dresden it was “scientific computing,” just as for Comrie in London.

The field opening itself to Van Wijngaarden’s observations in the winter of 1946, went under a variety of names and evolved in all directions. Douglas Hartree, whom he visited at the University of Manchester, claimed a new understanding of the field as “Numerical Analysis.” Hartree went around the world lecturing to spread the word.

Some things, by contrast, were not open to the visitors’ observation. For example, the cryptographic work done by Turing and others at Bletchley Park was only revealed in the 1970s. Also, even though less secretive, not every detail of cooperation in the field of computing between the UK and the US was shared.

While Van Wijngaarden was abroad. Mathematicians in the Netherlands got themselves organized for systematic research. On February 11, 1946, in Amsterdam they founded the Mathematical Center. With an eye on combining fundamental research and service, the Center was planned to have departments of Pure Mathematics, of Applied Mathematics, and for the service a Statistical Department and a Computing Department. Aad van Wijngaarden was soon convinced to change jobs from the aeronautical research laboratory to heading the computing department of the new Mathematical Center, as of January 1, 1947.

“It was the offer of freedom. Being young one can take a gamble. And most importantly, I got the opportunity to travel.”

Thus, he continued his reconnaissance tour, again to the United Kingdom in spring and to the United States during the fall of 1947. Over the summer he hosted a lecture by Douglas Hartree on numerical analysis and hired two assistants to start building a computing machine, a differential analyzer, during his absence. Building a differential analyzer was certainly a mark of Hartree's influence, but the trip to the United States had yet to begin.

The U.S. brought further novelty. Van Wijngaarden first stopped at Cambridge (Massachussets) to work with Howard Aiken at Harvard, next at Princeton, where Julian Bigelow was in charge of the construction of a computer supervised by John von Neumann.

From the 1930s at several spots in the US people had been constructing computing machinery. Analog devices like Vannevar Bush's differential analyzer - Van Wijngaarden had observed Hartree’s version of it, in Manchester - did not trigger his great interest.

Instead, he was soon captured by the greater advances made in the tradition of computation schemes. Originally such scientific computing had been slow computing using hand or electric calculators, such as Van Wijngaarden's own Marchant machine. John Atanasoff, George Stibitz and Howard Aiken, each in their own fashion, had been seeking ways to speed up the cranking of numbers. Their efforts were to create a new kind of machine, one that made a series of steps in a scheme, a sequence calculator. Howard Aiken, with support from IBM, was the most successful. He had a machine built and subsequently installed at Harvard in 1943, the ASCC, Automatic Sequence Controlled Calculator, otherwise known as Harvard Mark I. Sequences of instructions were in fact fed to the Mark I as paper strips.

Yet another machine had been created in Philadelphia. From 1943 through 1946, at the Moore School of Electrical Engineering a team headed by Presper Eckert, John Mauchly and Arthur Burks engaged in the automation of the production of gunning tables for the US army. In wartime new guns were invented and each new gun came with a mathematical table to calculate the trajectories of bullets shot from that gun. Thus far, the gunning tables had basically been hand calculated, but the production of tables could not keep up with the invention of new guns. Therefore, the team gained support to build the ENIAC, Electronic Numerical Integrator and Calculator. The name was telling: in fact they were hired to develop an electronic version of an analog machine and a computing device. Enter: John von Neumann, head of the Institute of Advanced Studies in Princeton. He and his aid Herman Goldstine got in touch with the Moore School ENIAC developers. Their meeting was not quite incidental because Von Neumann was always on the look-out for computing facilities in view of the Manhattan Project, which he advised. They sat together and new ideas spun off for a future machine, beyond ENIAC, a machine that would eat sequences of operations and execute them. It implied a new perspective, a view in which the sequence became a coded sequence. Von Neumann, with his habit of immediately jotting down his thoughts on paper, wrote a first sketch of a report, ‘First draft of a report on the EDVAC‘, June 1945. In full EDVAC was Electronic Discrete Variable Automatic Computer. The report always remained a draft because none of those involved found the time to finalize it. And others soon ran off with the ideas implied in the draft report. Von Neumann and Goldstine themselves wrote a series of further reports on the possibilities of creating a machine like the EDVAC, and fathom its capacities.

In the summer of 1946 ENIAC was completed. The war was over and with it the urgence of computer construction. To conclude the project, the Moore School organized a summer school in 1946. During that summer school the draft report was released. From that point on a dozen or so research groups continued on the ideas of Von Neumann’s First draft, commonly called “EDVAC-like” machines, including the Princeton group with Bigelow; including Moore School group that moved its actual ENIAC from Philadelphia to the Aberdeen Proving Ground in 1948 and had it adapted it to EDVAC-ideas; and prominently including Maurice Wilkes at the Mathematical Laboratory in Cambridge, UK. Wilkes’s EDSAC, Electronic Delay Storage Automatic Calculator, dedicated May 1949, is generally considered to be the first full realization of Von Neumann's concepts as expressed in the draft report. He named it EDSAC as a tribute to EDVAC. From the schemes of computation of scientific computing, two major leaps had been made. First, to elaborate computing procedures into sequences of computation. Next to turn them into coded sequences, “sequences of instructions” held in “memory”. The notion of “program”, let alone “programming”, had not been thought of yet. It arose in the actual practice of using the ENIAC. At hindsight, “EDVAC-like”, the result of the two conceptual leaps came to be coined “stored-program computer”. By itself, the expression “stored program” was first used in 1949 by an IBM research team at Poughkeepsie.

Aad van Wijngaarden arrived a year late for the Moore School summer school of 1946. He did meet with Wilkes in early 1947, who had attended. He joined Howard Aiken in Harvard in the fall of 1947. Aiken had meanwhile staged his own symposium in January 1947 on Large-Scale Digital Calculating Machinery. Aiken added a sobering advice, insisting on the use of well-proven technologies. Combining the technologies into a new machine would be enough of a challenge. Harvard Mark I and Harvard Mark II operated on sequences. Harvard Mark III was, in fact, an accessory machine to translate the schemes into sequences of instructions.

By contrast, in Princeton at the final stage of Van Wijngaarden’s tour, Bigelow submerged the European visitor in a much more ambitious and experimental version of the EDVAC-like machine.

Upon his return to Amsterdam in late 1947, Van Wijngaarden asked his assistants to drop the work on the differential analyzer in favor of the two lines of development towards an automatic computer which he just picked up, ARRA and AERA. With an uncharacteristic touch of modesty he projected a relay computer, for the short term. The ARRA, Automatische RelaisRekenmachine Amsterdam, was to follow the design of the ARC, Automatic Relais Computer, by Andrew Booth in London. An electronic computer, distantly inspired by Wilkes’s EDSAC, was projected for the long term. The plan went under the name AERA, Automatische Electronische Rekenmachine Amsterdam. In the 1940s the crucial characteristic of innovation was thought to be on the machine side. Conferences were called to discuss the new calculating machines. Accordingly, the community named itself ACM, the Association for Computing Machinery. Likewise, Van Wijngaarden served an IUTAM committee installed to compose a survey of existing machines. His initial publications were on machinery, but soon he shifted attention to the activity in A course on modern methods of computing. He would use the phrases "numerical analysis" and "scientific computing" interchangeably, adding the twist of qualifying them as modern. He taught programming, also with an emphasis on the activity, programming of automatic computers, first by himself, the second edition together with Edsger Dijkstra, and the third edition with Dirk Dekker and Dijkstra. The focus on the practice of computing was a constant in his work. It reached from building up a team of human computers practicing numerical analysis at a high level, via the creation of a specialist education in scientific computing, all the way to a recurring reflection on his part on “numerical analysis as a discipline.” Van Wijngaarden was very present as a head of department. Reflections on the discipline soon included reflections on programming, reflections on the nature of programming, reflections on the nature of applied mathematics. He lived the uncertainty of it being a scientific discipline in its own right, or not. He claimed his position and filled it. He did, meanwhile, assume an extraordinary professorship. He oversaw the creation of a computer manufacturing company, a spin-off from the Mathematical Center. He grew to become the acting director, and later full director, of the Mathematical Center. But at the core of it all was the creation of an academic field. Aad van Wijngaarden was very present in the European and international community, creating that academic field. One time he sailed with the Applied Mechanics community of IUTAM, another time on board with the mathematicians, then again he straightforwardly engaged in a discussion on computing machinery and human thought. During the 1950s he traveled and traveled again from Sweden to Turkey, giving lectures, stopping in Rome, stopping in Santa Monica, attending a week-long conference in Darmstadt.

The joint conference of communication engineers, and applied mechanics scholars, NTV-GaMM, in Darmstadt, October 1955, brought together the larger community for the first time since the Second World War. Everybody attended. From the U.S. Howard Aiken, Alston Householder and Herman Goldstine, gave keynote lectures. Because the Russian delegation could not make it in time, the meeting was simply extended by a day. It had the effect of changing the very landscape of the field under construction. In name yet another one of these conferences on machinery, there was a whole section devoted to ways of programming. In the debate Heinz Rutishauser, of Zürich, called for a unified notation of programs. His plea set in motion a whole series of events, which, taken together, call into question whether “numerical analysis” would still be a fitting term to describe a paradigm of working with computers. For one thing, next to the words "scheme" and "procedure," the word "algorithm" was gaining currency. For another, from the sideline came another word, namely "language" for the system of describing, or rather expressing, instructions. Perhaps these new expressions signified that the field was no longer “numerical analysis.”

Within Germany a small committee of numerical analysts got together to discuss such a unified notation. Rutishauser from Switzerland, Friedrich Bauer and Klaus Samelson from Munich, N.J. Lehmann from Dresden, and Bottenbruch from Darmstadt. Simultaneously Alston Householder, the chair of the ACM, invited Friedrich Bauer and Hermann Bottenbruch over for a visit to the United States. The ensuing trip brought them into contact with Alan Perlis and John Carr, two who had led the way in using the word "language" for programming systems. It led to a further invitation to create a joint the committee of ACM and GaMM, i.e. the German applied mechanics community, to discuss a unified or common way of programming automatic computers. It aimed for what was then called by a relatively new expression: a programming language. Rutishausers plea for a unified notation had thereby turned into a search for a common programming language. In Zürich a first draft of such a programming language was agreed upon. At first it was called IAL, International Algebraic Language. At second thought, following a suggestion by Bottenbruch, it was renamed ALGOL, for ALGOrithmic Language. At this point Van Wijngaarden stepped in, finding the expression “programming language” on the doorstep.

Search of new purpose was what brought Adriaan van Wijngaarden to that doorstep.

The invitation to address the International Congress of Mathematicians in Edinburgh, August 1958, had certainly been a jewel in his crown. On August 18 he gave his talk on ‘Summation of series,’ contributing to the very core of numerical analysis, then moved on to spend holidays in Scotland with his wife. Aad van Wijngaarden and his wife Babs were involved in a car crash. His wife died and he was seriously injured, hospitalized for three full months. He was devastated.

Seeking a road to recovery, he made a deliberate new start. Some of his closest friends shared the sad experience of being put at a distance; seeing their friendship frozen. Now specializing in programming languages, Aad sought a new subject and new company.

Van Wijngaarden had always kept a fascination for language, but up to this point had left the reflection on programming largely to Edsger Dijkstra. Specializing in programming languages, he joined the very next meeting in February 1959 in Copenhagen. At the following meeting in Mainz in July, Friedrich Bauer expressed his delight with the new participation. Aad van Wijngaarden returned the compliment and added that now that they were working so closely together, they might be less formal: “My name is Aad.” Clearly, this was his new company.

The ACM-GaMM committee was proceeding towards an agreement on ALGOL at a high pace. The increased informality was perhaps helpful; the main factor was certainly the manner in which Peter Naur in Copenhagen took minutes and circulated the ALGOL Bulletin. Commissoned to take minutes he was the de facto editor.

ALGOL in Parijs, late1959

Far right Friedrich Bauer

Second right Klaus Samelson

Fourth right Edsger Dijkstra;

achter hem op de derde rij Aad van Wijngaarden.

Further meetings in Paris in November and in Mainz in December prepared for the ALGOL 60 meeting in Paris in January 1960. Amidst all of this, Van Wijngaarden and Dijkstra had their own agenda. For them it was a matter of beauty and a matter of decency that ALGOL, allow the most general of constructions to define procedures; it should allow recursive procedures. As a general principle this had been voted down in Paris. They did not give up. In the next month Van Wijngaarden, seconded by Dijkstra, called Peter Naur on the telephone. They proposed a text that spelled out the acceptance of recursive procedures --without mentioning the notion of recursion. Naur, in turn, consulted Friedrich Bauer. Bauer clearly understood the phrase and understood that he had lost the play on words. He agreed, and that was the acceptance of the recursion in ALGOL 60:

ALGOL 60 was published and met with enthusiasm. It was accepted in the broader community of working with programming languages. By itself it became the standard of what was since called a higher-level programming language.

From that point on three simultaneous developments took place: the implementation of ALGOL 60; the founding of IFIP; the adoption of ALGOL by IFIP.

One may replace programming systems with programming languages, but that does by itself yield a working system, an operating system. Creating such a system would be called “implementation”, sometimes also a “translation”. In other words, creating a programming system based on ALGOL 60 was the implementation of ALGOL 60. On their next visit to Copenhagen in March 1960, Edsger Dijkstra, Jaap Zonneveld, Freek Barning, and Aad van Wijngaarden discussed the task of making an implementation of ALGOL 60. They self-imposed a challenge that they would do so in 6 months and challenged the team in Copenhagen to come and see them in August and observe the implementation of ALGOL in operation. Dijkstra and Zonneveld did the work. They were well on time, but Naur had not taken their challenge seriously and did not show up in Amsterdam. The international ALGOL community was baffled. They delivered a transparent and beautiful product. Their British colleague Brian Randell noted that a new standard had been set for the writing of implementations.

In parallel to the growth of an informal ALGOL community, a more formal grouping originated in the same year, in the same city, and with many of the same participants. Isaac Auerbach had been chairing a committee of the professional societies in the U.S. aiming to spread their wings internationally. Their aspiration was organize an international conference on information processing. Once appointed U.S. delegate to UNESCO, Auerbach carried with him to Paris the initiative to organize. The International Conference on Information Processing was held in June, 1959. Aiken, Householder, Wilkes, Walther, they all participated. Van Wijngaarden came on special invitation and was awarded the silver medal of the city of Paris. The overt initiative concealed another, namely, to set up an international organization for information processing. That secondary agenda required more diplomacy than expected. Eastern European delegations were willing to join only if participation was restricted to one society per country; no individual members. The result was a federation, the International Federation of Information Processing Societies, IFIPS. Later, in 1961, the S was dropped from the name, IFIP. The federation was founded with 13 members on January 1, 1960. Thanks to Van Wijngaarden, the Netherlands were in.

In June 1958, while preparing the conference, he had realized that the Netherlands didn't even have a society that could participate in a federation. Just in time for IFIP the Nederlands Rekenmachinegenootschap was founded, NRMG holding its 0th meeting on April 25, 1959.

The Federation existed, but only barely. Two years on, a further congress was convened in Munich, by the success of which IFIP proved its viability. It got itself organized into TC’s, technical committees and these TC’s each had several working groups, the most important of which was WG 2.1 on Algorithmic Languages. At this point the two streams of events merged. ALGOL was now a matter for IFIP and the ALGOL committee was now an IFIP working group.

ALGOL 60 saw a revised report issued in 1962. WG 2.1 saw as its task the maintenance of that language. The discussions were whether the working group would be developing a new language, a successor, or restrict itself to updating the given definition. Change of status came with change of personnel. Being part of the IFIP organization, the same people gathered under a different regime. Peter Naur stepped aside, dissatisfied as he was with the ALGOL group now being a formalized organization; no good could come out of that, he considered. Friedrich Bauer, who had been the informal leader of ALGOL, was happy to leave the chair of Working Group 2.1 to a Dutchman, Willem van der Poel.

Working Group 2.1 convened with increasing frequency. Van Wijngaarden traveled and traveled. Visiting professor at Berkeley, he met the bright young student from Switzerland, Niklas Wirth, whom he invited to attend the discussions in the working group. At their next meeting in Munich, with a play on the controversy in these discussions over “calling by value” versus “calling by name”, he introduced Wirth, saying, "You may call him by value, you may call him by name."

Van Wijngaarden, who had been of one mind with Dijkstra in matters of recursion, was now eagerly pushing the boundaries even further. Continuing the path of abstraction, he preferred to see no limitations, no boundaries to what one could design as a new programming language, no boundaries to what one could express in a new language. One can see him trying to reach elegance and generality, Generalized ALGOL in 1962, Recursive definition of syntax and semantics, 1964. That effort to elevate recursion to a higher level, left even Dijkstra “somewhat baffled”. One can see the Working Group trying to keep Aad on board, by differentiating between ALGOL X containing suggestions for an immediate successor language versus ALGOL Y for a more distant future. That is how they parted in Princeton in May 1965 with Gerhard Seegmüller, Niklas Wirth and Aad van Wijngaarden taking homework for the next meeting in Saint-Pierre-de-Chartreuse in October. Seegmüller delivered his proposals well on time. Wirth spent the summer in Amsterdam to develop his ideas on a successor language in discussion with Van Wijngaarden. Instead of leaving it at that, Van Wijngaarden, at the very last moment, completed his draft, for an ALGOL X (not Y), rudely overtaking his fellow authors. Report MR 76 was distributed on the spot in the French mountains: Orthogonal design and description of a formal language. It was his day of triumph. Against all odds, against all previous criticisms, he convinced the working group with his two-level grammars.

He had found a way to avoid critique and dismay. The approach of two-level grammars was so attractive that the working group decided there and then that whatever would be their final proposal for a successor to ALGOL 60, it should be phrased in Van Wijngaarden's fashion. Wirth held no grudge for being overtaken. Seegmüller realized that, in fact the working group had been longing for such abstract tools in order to make a bigger leap in programming languages.

It took another 3 years before the proposal for the next language was discussed and agreed upon. In that sense the late arrival in Saint-Pierre de Chartreuse was just a prelude to a succession of missed deadlines and sustained disagreements. Successive versions appeared as A draft proposal for the algorithmic language ALGOL X; A draft proposal for the algorithmic language ALGOL 67; A draft proposal for ALGOL 68; A draft report on the algorithmic language ALGOL 68; Working document on the algorithmic language ALGOL 68; Penultimate draft report on the algorithmic language ALGOL 68; Final draft report on the algorithmic language ALGOL 68.

ALGOL 68 was presented December 1968 at the IFIP Congress, in Edinburgh. Report on the algorithmic language ALGOL 68 appeared as Mathematical Center report MR 101 in 1969. Hiding on the back cover the familiar @ symbol.

Van Wijngaarden could not stop himself from further musings on the two-level theme: On the generative power of two-level grammars (1974), Thinking on two levels (1979). In a final liberation from language he showed, in full appreciation of the implicit irony, that using two-level grammars one need not even define a particular language Languageless programming (1981).