The Last Thing They Did Together

The carpool from the IBM Research Laboratory in San Jose was routine. Don Chamberlin drove that day in June 1974, the same route he’d taken countless times with his best friend and colleague Raymond Boyce. They were not just coworkers—they were neighbors in the suburban sprawl south of the Bay, the kind of friends who built their lives in proximity, who debated query languages over morning coffee and Saturday dinners. They’d moved to California together, left the East Coast behind, and thrown their considerable minds at a single problem: how to make Edgar Codd’s theoretical revolution into something real, something practical, something that ordinary people could actually use.

By lunchtime that day, Boyce had collapsed in the cafeteria. A brain aneurysm. No warning. No goodbye. He died on June 18, 1974, leaving behind his wife Sandy and their infant daughter Kristin. He was twenty-seven years old. He and Chamberlin had just published their first SEQUEL paper at the ACM SIGFIDET conference in Ann Arbor, Michigan, less than a month earlier. That paper—delivered in May, already distributed, already being read by the people who would build the future—became the last thing they did together.

This is not a story about a single visionary genius. It’s a story about two people who understood something that almost everyone else, including IBM itself, did not yet see.

In 1970, Edgar Codd, a British RAF veteran who had served as a pilot during the Second World War, published a paper that was destined to change everything. “A Relational Model of Data for Large Shared Data Banks” appeared in Communications of the ACM. Codd had moved to IBM’s San Jose Research Laboratory in 1968 and spent the next two years crystallizing decades of thought into eleven pages that would alter the course of computing. He proposed that data could be organized into simple tables, that these tables could be related to one another through common values, and that this organization could be described with mathematical precision. It was elegant. It was revolutionary. It was, in the eyes of the corporation paying for it, not ready for the market.

IBM had a successful database product called IMS, and was not in a hurry to replace it. Instead of rushing Codd’s ideas to market, IBM organized a research project, called System R, to investigate the feasibility of relational databases. The research project was allowed, even encouraged, to publish its work in the open technical literature. A professor at Berkeley named Michael Stonebraker would later wonder aloud why IBM, having invented the relational model, had taken so long to commercialize it.

In the summer of 1973, Don Chamberlin and Raymond Boyce arrived in San Jose to work on the System R project. Their mandate was to implement Codd’s model. They needed a language—not a language for mathematicians writing predicate calculus, but a language for ordinary people, for database administrators and office workers and anyone who simply wanted to ask the database a question. They called it SEQUEL: Structured English Query Language. It read almost like English. A person could learn it. A person without a PhD in mathematics could sit down and write a query.

The two men had developed their ideas by playing what they called “the query game.” Chamberlin would dream up a question he wanted to ask of a database, and Boyce would write the SEQUEL code to answer it. Then they’d reverse roles. Question, answer. Answer, question. Back and forth, refining the language, simplifying it, testing the boundaries of what could be expressed, learning from each attempt what the language still needed. This wasn’t academic exercise. This was craftsmanship. This was two people treating a programming language as something that should serve human beings, not the other way around.

In early May 1974, in Ann Arbor, Michigan, they presented “SEQUEL: A Structured English Query Language” to the ACM SIGFIDET workshop. The paper was straightforward, practical, almost modest in tone. Here is what we built. Here is how it works. Here is why we think people should use it instead of wrestling with complex operations. Within a month, Boyce was dead. Chamberlin, who had relied on his friend for the intellectual partnership that had powered their work, had to absorb the loss and continue alone.

But there was a young entrepreneur paying attention. Larry Ellison read the System R papers and correctly predicted that IBM would eventually release a SEQUEL database on its mainframe computers. Seeing a market opportunity, Ellison built his own SEQUEL database on a less expensive minicomputer.

Ellison’s product, which he named Oracle, beat IBM to the relational database market by about two years.

The language itself would not remain unchanged. Trademark disputes—a company called Hawker Siddeley owned the rights to SEQUEL—forced a renaming. They dropped the vowels. SEQUEL became SQL. The acronym stood for Structured Query Language now, not Structured English Query Language, and though Chamberlin would spend the rest of his life pronouncing each of those three letters when others said it as a word, that change was pragmatic and necessary. It had to survive. The language had to survive.

And it did. Chamberlin remained at IBM, elected eventually to the National Academy of Engineering. He watched as SQL became the most widely used database query programming language on Earth. Every day, more queries are, not according to every measure and not at every moment, but by some counts yes—more queries executed in SQL than in any other language, across more machines, more systems, more databases than any other framework for asking questions of stored information. Banks run on it. Hospitals run on it. Airlines run on it. Every purchase you make online runs through SQL. Every balance inquiry, every reservation, every transaction that matters to modern life has probably been processed by SQL code.

Boyce did not live to see this. But his name endured. While working on the theoretical foundations of relational databases, he had independently articulated what became known as Boyce-Codd Normal Form—a mathematical tool for designing databases properly, for eliminating redundancy, for ensuring data integrity. It bears his name still. In database schema design classes all over the world, students learn about BCNF, and they’re learning the contribution of a man who died at twenty-seven and never saw the full impact of his work.

There is something almost unbearable about this. Two friends, neighbors, carpool partners, devoted to making something powerful also usable. One of them dies after shipping the thing they built together. The other continues in solitude, carrying the work forward, carrying the loss, watching as the language they created together becomes central to electronic commerce, the foundation of everything. That final paper, delivered together in May 1974, stands as the moment when they were both still here, still building, still believing that clarity and accessibility could be expressed through mathematics and code.

SQL was not the only language born from that era. But it may be the only one that became truly indispensable. Not because it was the most elegant. Not because it won some theoretical competition. It became indispensable because Chamberlin and Boyce built it for human beings instead of for computers. They asked a fundamental question: how would a person naturally want to ask a database for information? And then they built a language around the answer.

The language outlasted them both, though Chamberlin lived to see it triumph and Boyce did not. Today, fifty years later, when a database administrator needs to retrieve data, when a data scientist needs to explore a dataset, when a company needs to query the millions of customer records on which its business depends, they use SQL. They use the last thing Don Chamberlin and Raymond Boyce did together. They use it dozens of times a day. They use it without thinking about the two friends in California who drove to work together and changed everything, or the one who came home alone.