The first article in "The Software Snarl" series {Dec. 9-12} was too amusing {"Mass Production Comes to Software; Japanese Seek to Challenge U.S. Lead by Simplifying Process"}. In the illustrated example at the top of page A24, for example, the programming problem was that "the programmer neglected to instruct the computer that the cake would be baked in Denver, where special high-altitude flour would be needed."

But did the chef tell the programmer, and did the programmer simply ignore the chef? Not likely. That's a communication breakdown that has little or nothing to do with programming. However, it is in fact typical. Most communication problems in any program involving software are blamed on programming.

About five years ago, I led a study of software costs and problems for Mandex Inc. sponsored by the U.S. Navy cost accounting office. We pointed out the above fact and a few other misconceptions. One of these misconceptions is that the Navy cost estimates for software development always prove to be accurate. That's true, but not for the reason you'd think. When money allocated for the development is spent, the software is almost always accepted -- no matter what condition it's in. The development is then completed under new funding called "software maintenance."

Another misconception is that the software requirements don't change during the development, even if, say, 10 years are required for it. This conclusion is based on the fact that very few requirement specification documents are changed during the development. The truth is that no one looks at these documents after the award of contract. The true requirements -- what the software is really supposed to do -- are communicated informally. By the time the government completes its paper work and signs the contract, the formal requirements document is already out of date.


After having worked the Advanced Field Artillery Tactical Data System as well as the other four-fifths of the Army's "Battlefield Automation (C3I)" requirement over most of the 1980s as a project engineer and senior staff manager at the Army Materiel Command, I believe "The Software Snarl" article of Dec. 11 {"Pentagon Finds High-Tech Projects Hard to Manage"} was somewhat misleading. Most of the government and industry people involved, regardless of their computer-programming acumen never doubted the true cost and time for development and deployment of the AFATDS and others.

The fact is, all major software-driven programs -- commercial or governmental -- evolve at their own pace, with lots of expensive "lessons learned" when the required initial planning and associated dedicated resources are not in place from the beginning. They are victims of our political and institutional shortsightedness, which makes liars of us all in our effort to make at least some progress and/or profits.

ROY L. ARNOLD Fairfax Station

In her four-part series, Evelyn Richards has masterfully described the growing importance of software in our society and the huge challenge faced by government and industry in developing large-scale software systems that perform as desired. Achieving the higher levels of productivity needed to meet this challenge means that the free-wheeling, "genius" programmers (read: hackers) of the past must be disciplined in the application of sound engineering principles when designing, testing and implementing large-scale software systems.

In addition to the commercial companies described in this series, which develop products to automate various portions of the software development process, several U.S. industry and government initiatives are forging methods and tools to apply proven engineering practices to the "black art" of software development. The Defense Department-sponsored Software Engineering Institute at Carnegie-Mellon University, the Microelectronics and Computer Technology consortium in Austin, Tex., and the Software Productivity Consortium, established by the nation's leading aerospace and defense companies in 1985, are examples.

Each of these groups actively communicates and collaborates with other government, commercial and academic organizations in advancing the concept of software engineering. Only through such close liaison and cooperation can we hope to prevent the loss of yet another American competitive edge.

ARTHUR I. HERSH President and Chief Executive Officer Software Productivity Consortium Herndon