Published on October 26, 2015, this first installment of a three-part retrospective follows the author’s first programming job. In 1999, after the dot-com crash, the Chicago suburbs offered few openings. Job seekers checked newspapers, and the author faxed resumes. A company seeking entry-level programmers revealed little about its technology. The interviewee brought ten pages of C++ examples that used OpenGL through GLUT. The interviewer warned that the work could become boring, but the author accepted the offer.

The job proved instructive. It taught the author analysis, project management, professional conduct, and customer relations. Pick still defined the first seven years of the author’s career, and leaving it cost about $1,000. During a later visit to the Computer History Museum in Mountain View, California, the author responded to a photograph of Pick’s creator with an obscene gesture.

Richard “Dick” Pick created the Generalized Information Retrieval Language System, known as GIRLS, in 1965 for the IBM System/360, which had launched in 1964. The controversial name disappeared before the commercial release of the Pick Operating System in 1973. Pick ran on mainframes and minicomputers. Later editions operated as applications on Windows and Unix, with behavior closer to an emulator than a virtual machine. The original system supported multiple users and time sharing. By the author’s time, it also provided virtual memory and could use up to one gigabyte of RAM on its host.

Pick combined operating-system functions with a database. Its early-2000s marketing described the model as post-relational. Files used hash-table structures with primary keys and free-form text data. Attribute mark ASCII 254 separated fields. Value mark ASCII 253 allowed one attribute to contain multiple values, including aligned associated multivalues. Subvalue mark ASCII 252 added another level, while one Pick variant used ASCII 251 as an additional delimiter. Records in the same file could therefore have entirely different shapes and could contain structures nested within structures.

Each file also had a dictionary stored in a secondary file. Its attributes, values, and subvalues described the data in the main file and supported queries. Pick enforced none of the described data types or other record rules, and one attribute could receive two names. The planned second article was to cover querying this unusual structure.