January 2007

Anything You Can Do, I Can Do Meta

Continued from page 3

By Scott Rosenberg

Audio »

Favorite

E-mail

The Machine's Language

Simonyi was born in Budapest in 1948. The son of a physics professor, he fell in love at 15 with his first computer--a mammoth Russian Ural II in Hungary's Central Statistical Office. By the 1960s, the Ural, which received its instructions through cash-register-style keys and had a roomful of vacuum tubes to perform calculations, would already have been a relic anywhere else in the world. But Hungary's Communist leaders were trying to use the Soviet castoff to optimize rail and trucking schedules. The Ural wasn't up to the task: there was no way to input real-time data on shipments. "It was completely hopeless," Simonyi recalls. "It could have been done very easily by supply and demand. Unfortunately, that was politically incorrect."

But Simonyi didn't care. "I loved that computer," he says, "even though it was useless." As a child he had built an Erector Set car with a four-speed transmission--not so much because he wanted to play with it as simply to understand how it worked. A former student of his father's found Simonyi a job as the Ural's night nurse. Because the machine blew out a tube each time it was turned off and on, the Statistics Office preferred to allow it to run all night. Thus, from dusk to dawn, the mainframe was all Simonyi's; he had a personal computer before such things existed. He learned to program it by writing clever but useless routines to generate "magic squares"--numerical arrays in which the sums of the rows, columns, and diagonals all match.

Programmers elsewhere in the world had already invented a Babel of programming languages--Fortran, Cobol, Lisp (a fabled language: see "Ancient Text," p. 20), and so on--to ease their work, which then as now consisted of painstakingly writing elaborate sets of instructions for computers to execute. In those languages, the instructions took the form of lines of text that were entered on keyboards and frequently stored on punch cards. This "source code" was then "compiled," or translated into "machine code"--the 1 s and 0 s that a digital computer could understand. The method remains largely unchanged today, even if most programmers now use programming tools running on ordinary PCs. But on the Ural, Simonyi learned to program at a more primitive level, laboriously punching in the "opcodes" of machine language, specifying, instruction by instruction, the sequences of memory fetches, additions, memory stores, and jumps that the computer's processor had to follow to execute even the most trivial operation. It was (as Simonyi told author Steve Lohr in the 2001 book Go To ) "Stone Age programming." Simonyi still remembers the codes. "Twenty-two is JUMP," he says today. "It's burned into my ROM."

Hungary in the 1960s, still flinching from the Soviet suppression of its 1956 revolt, was not a place for an ambitious young man with a taste for problem-solving. At 17, Simonyi landed an internship with a Danish computer company by showing some of its programmers samples of his hand-coded Ural programs. The Hungarian authorities expected Simonyi to return; he'd already won a coveted university spot. Instead, with his father's encouragement, he fled to the United States.

Anything You Can Do, I Can Do Meta

Would you like to read more articles from the January/February 2007 issue?

Comments

Related Articles

Current Issue

Magazine Services

Career Resources

MIT Technology Insider

More Technology News from

Wire Stories