Computerworld logo
(Print 08/11/97)

Cobol pioneer pitches year 2000 fix
Tim Ouellette and Robert L. Scheier

Bob Bemer has been around computers since the birth of the industry, and now he thinks he has a way to make sure the year 2000 isn't the end for many systems.

The industry veteran has come up with a plan to attack the date problem in the depths of the system the machine code of ones and zeros. He claims this would be cleaner than some current methods and up to 10 times faster, too.

But his product isn't shipping yet, and he has to move quickly to reach users before they turn to other year 2000 solutions. Furthermore, some observers are skeptical about the product.

Bemer, 77, had a hand in the creation of Cobol and the adoption of the ASCII naming standard. After a career at IBM, Univac, Rand Corp. and General Electric Co., Bemer founded BMR Software, Inc. in Dallas to market his new product Vertex 2000.

"With the problem focused on Cobol, I felt a personal responsibility for the whole thing," Bemer said.

Vertex 2000 examines a mainframe program's object code which is structured machine code finds every possible date instance during an off-line scan and patches the code to run a separate subroutine for handling the date. Then when the program runs, the subroutine uses extra bits in a date field, which Bemer calls ``Bigits,'' to indicate the century.

Other automated tools on the market take a similar approach at the higher-level source code, which requires more time and manual intervention to get the job done, Bemer said.

The problem, Bemer said, is ``who's going to look at [the source code]? Not the guy who wrote it because he is most likely gone.''

He also said the method eliminates the need for testing, which can take up to 50% of a year 2000 project's time and effort. On the other hand, he admitted that his program will take a performance hit when the user runs applications because it is correcting data instances online.

"Our program will run slower to begin with but soup itself up after it weeds itself of all the things that weren't year operations, he said.

Some observers and year 2000 practitioners said that even if Bemer's approach works, it tackles only code conversion, which is the easiest and cheapest part of a year 2000 fix.

"It's only addressing 20% of the cost of the problem," said University of North Texas professor Leon A. Kappelman, co-chairman of the Society for Information Management's year 2000 working group. "That's not trivial, but it's not a silver bullet in the sense you wave your magic wand and everything is fixed."

Assessment, project management and testing make up the bulk of year 2000 efforts. So just understanding how to fix code at the object level, he added, ``doesn't necessarily mean you understand the complexity of enterprise systems and understand the requirements of testing.''

A number of posters on year 2000 Internet discussion groups have debated Bemer's plan. Many voiced concern over altering low-level object code and wondered whether he actually had enough time to turn his idea into a workable product for widespread use.

BMR Software won't have Vertex 2000 ready to attack actual customer code for another two months.


Copyright © 1999 Computerworld, Inc. All rights reserved.