Aiming a 'silver-plated bullet' at year 2000 problem
By Thomas Petzinger Jr. / The Wall Street Journal
PHOENIX - On the upper-left corner of your computer keyboard is a button labeled "esc." Behind that key is a 37-year-old process called the "escape sequence," created by a former IBM programming executive named Bob Bemer.
Today, at age 77, Bemer has invented another escape routine: for the vexing year 2000 computer crisis. Bemer's solution won't remedy everybody's problem, and the boldness of his method terrifies many. But even if his idea falls short of the hoped-for "silver bullet" for the millennium bug, it is, he says, a "silver-plated bullet."
As everyone in business should know, mainframe computers (and many desktop systems) will go berserk after Dec. 31, 1999, (and some sooner than that) when the longstanding practice of using two-digit years will cause computers to treat "00" as 1900. Because every language and application is unique, fixing the problem involves tedious analysis - more work than programmers can likely complete in the next 30 months. Automated tools exist, but all require human hand-holding.
Enter Mr. Bemer - perhaps the only living programmer, for reasons you'll see, with the knowledge
necessary to automate the fix.
A wartime mathematician for Douglas Aircraft and later a Hollywood set designer, he saw his first computer in 1949 and never looked back, working for IBM, Univac, GE and others. As IBM's chief of programming standards, his creation of the escape sequence in 1960 allowed computers to break from one alphabet to another, a critical step toward laser printers and cursor movement. He led the effort to establish a universal character set known to millions as ASCII. He created the name COBOL for what still ranks among the world's dominant computer languages. He helped develop the standard by which binary digits (bits) travel in packs of eight (bytes). He even created the backslash.
Writing in Interface Age in 1979, he warned that the year 2000 would cause big problems, but after retiring in 1982 he pretty much forgot the issue. Then, two years ago, he read a Front Lines column about David Eddy, a software marketer pleading with corporate America to address the issue. Though already a great-grandfather, Bemer swung into action.
The best solution, he thought, existed at the most fundamental level, where computers operate in ones and zeros through something called object code (or machine code). The myriad programming lan
guages that now exist were made possible by a decades-old software innovation called a compiler, which translates higher-level languages into the arcane code that actually runs the machines. Today's programmers don't even think about the object code, but Bemer knows it cold. Attacking the problem at this basic level would avoid the need for thousands of individual solutions at a higher level.
His method examines every arithmetic operation running through object code. Along the way, special rules identify those operations involving dates. (Dates are never multiplied or divided, for instance.) At that point a variation on Bemer's old escape sequence comes into play, detouring the date, as if along a railroad siding, to a place where the missing century is hooked to the two-digit year. "Everyone else is trying to find all the years," he says. "We let the years find us."
Rehabilitated, the date then returns to the application. But the millennium problem is so maddening partly because many systems have no room to cram in the extra digits representing centuries. Bemer gets around that by using spare bits within the year bytes - "four pounds in a two-pound bag."
He also proposes to create additional numeral sets - one with overlines, another with u
nderlines - to indicate which century they belong to when printed out or displayed. (He has contacted the necessary standard-making bodies seeking new fonts.) He admits that where programmers used shortcuts, such as using "99" to represent the year infinity, his solution will fall short.
Bemer's concept is immensely more complex than I could ever completely grasp, much less fully describe in three paragraphs. But he permitted me to share some of his patent materials with a number of leading 2000 authorities I trust. Most agreed his concept would work, and much faster than existing tools - but many called it too radical. The idea of tampering with source code and stored data makes some people apoplectic.
But the authorities agreed that as time runs out on more conservative methods, such misgivings may diminish. "People may not feel real comfortable with it - but they'll be less comfortable with the failure of their businesses," said Joanne Metta-Sullivan, who is attacking the problem at Merrill Lynch. Likewise, Electronic Data Systems Corp. studied Bemer's concept and wrote him a letter saying his invention could "offer many customers what they will need to solve their year 2000 problem, especially those who wait until the last minute."
must still turn his concept into working code. He's moving to Dallas to work with two partners there: Systems Source Inc. and Millennium Consulting Group. He plans to finish this summer, although as everybody knows, software projects are seldom completed on time these days.
But it's probably unwise to judge Bemer by today's axioms. "Back then," he says, "we weren't blinded by conventional wisdom."