........... COMPUTER SCIENCE

MILLENNIUM BUG ZAPPER

A radical solution for the Year 2000 problem

Like the late Jonas Salk's return to prominence to combat HIV, Bob Bemer's emergence from semiretirement to solve the "Year 2000" computer problem evokes nostalgia. A pioneer in the digital world, Bemer is the man who, among other accomplishments, helped to define ASCII characteristics, which allow otherwise incompatible computers to exchange text. But critics say Bemer's solution, though ingenious, may be too much too late.

The Year 2000, or "Y2K," problem arises from the widespread use of two-digit date fields, which leaves computers confused whether "00" refers to the year 1900 or 2000. According to detailed estimates, fixing this pervasive "bug" will cost companies and governments around the world hundreds of billions of dollars. One brute-force solution calls for finding every instance of a two-digit year and then rewriting the computer code to expand each field to accommodate four digits. This hellishly tedious process has led a cottage industry of vendors to create software tools that help to automate the procedure.

The unique feature of Vertex 2000, Bemer's innovative solution, is that it expands dates "vertically." Bemer realized that each space reserved for two-digit years contains surplus data bits. This tiny excess was enough room to piggyback additional information for denoting the century. He dubbed the vertically expanded characters "bigits," for Bemer digits. This efficient approach offers a huge advantage over the conventional method of "horizontal" expansion (that is, going to four-digit years), which leads to longer files and, as a result, possible computer crashes. How did Bemer know that he could squeeze the century information into existing two-digit years? "I've been in the character-set business since 1960," he declares. "I've eaten, slept, breathed and lived character sets."

To incorporate these new bigits, Vertex 2000 makes the necessary changes in object code, or machine language, which a computer (but not humans) can understand easily. Traditional methods work in the more accessible world of source code, which is written in computer languages such as COBOL and FORTRAN that have comprehensible commands such as "write" and "read."

Because Vertex 2000 is supposed to do much of this conversion inconspicuously and automatically, Bemer asserts that his method will be at least 10 times faster than other solutions, at less than half the cost. To capitalize on these possible benefits, he founded BMR Software in Richardson, Tex. Expecting business to ramp up soon, Bemer says, "We're sitting here in a state of delirious excitement."

Others are also delirious but not with excitement. Making changes at the arcane level of object code gives many programmers conniptions, and not everyone is crazy over bigits. "Bemer has got to be respected for what he's done, but his approach is so advanced in theory that you have to question whether it's practical," asserts Leland G. Freeman, consultant with Management Support Technology in Framingham, Mass. But Bemer contends that extreme approaches are exactly what companies need right now. "It's either tamper with your object code or have your business go belly up," he argues.

Bemer does, however, acknowledge that Vertex 2000 is merely a temporary fix. Adjusted programs will typically run about 20 percent slower. Consequently, Bemer states that companies should use his method to buy time before implementing a permanent solution, such as a conversion to a Julian Day system.

That said, BMR Software had better move fast. At press time, the company had yet to ship Vertex 2000 even after having pushed back the schedule several times already. With only a year and a half left until the new millennium, Freeman says, "by the time Bemer's solution is accepted in the market, the game could be over."

--Alden M. Hayashi