Information Processing

30,000 Programmers Later, 'Software Gridlock'

IBM promised a way to link all of its machines. That was 1987

Evan I. Schwartz in New York, with bureau reports
Business Week

July 15, 1991

Flash back to 1986. IBM had entered a deep slump. Sales inched up only 2%, while profits dropped 27%. Meanwhile, chief rival Digital Equipment Corp. was having a field day, rightly claiming ''Digital has it now'' -- the ''it'' being a range of computers that all run the same software and easily work together in networks. DEC's net profits soared 115% that year, as revenues grew 20%, to $ 8.4 billion.

So in March, 1987, IBM responded, announcing Systems Application Architecture, an ambitious plan for software to weave together IBM's incompatible computers. If they used software that fits the SAA scheme, IBM assured its frustrated customers, within a few years they would be able to share programs and shuttle information easily among mainframes, minicomputers, and desktop systems -- capabilities that even DEC couldn't match.

TIRED OF WAITING

But SAA was more than IBM's response to a surging rival. It became the cornerstone of Big Blue's strategy to keep mainframes -- which brought in the bulk of its sales and profits -- at the center of corporate networks. If SAA could provide easy access to the mainframe, customers would be less likely to shift control of their networks to increasingly capable minicomputers and PCs. IBM also saw that by the 1990s, rapid growth and high profits in computers would come not from hardware but from software and services. Supplying pieces of SAA would be a way for IBM to capitalize on that trend, too.

Flash forward to today: IBM's customers still don't have the smooth links SAA promised, and many are tired of waiting. ''SAA is late,'' says Jack L. Hancock, executive vice-president for technology at Pacific Bell. Now, he is looking for other ways to do what IBM promised with SAA. ''It represents a loss of revenue and ego on IBM's part,'' he adds.

Indeed, one cause of the declines in market share and earnings that have enraged Chairman John F. Akers is IBM's slowness in delivering software that would bring the SAA concept to life. ''Software gridlock is one of the roots of the financial issues,'' says Michael Braude, head of research at Stamford (Conn.) consultant Gartner Group Inc.

From the beginning, IBM knew that SAA was a Herculean undertaking. It required protocols and programs to pave over the disparities among IBM's many different computers. Eventually, an estimated 30,000 programmers would be assigned to the task. IBM made a list of languages and tools to be used in building SAA programs and named Earl F. Wheeler SAA czar -- to make sure that IBM's programming efforts stayed focused on the project. To encourage outsiders to support SAA, IBM spent $ 500 million to buy minority stakes in software and service companies.

The initial concept was admittedly sketchy. But in May, 1989, IBM showed a concrete example of how SAA could work. It introduced OfficeVision, a series of programs that ran ''cooperatively'' on PCs and IBM mainframes and did such things as manage electronic mail and extract data from a mainframe for use in a PC spreadsheet.

But OfficeVision has become a symbol of SAA's shortcomings. The most important part -- software linking PCs and mainframes -- was due in March, 1990. It has been postponed twice, and IBM no longer even says when it will be ready. ''I've already written off OfficeVision because it has been delayed so many times,'' says Jon W. d'Alessio, chief information officer for McKesson Corp., a San Francisco pharmaceuticals supplier.

To keep other customers from bolting, IBM is forging deals with independent software makers. On June 24, it said it will sell Lotus Development Corp.'s electronic-mail program and Notes, a program that lets workers on a network coordinate tasks. Earlier, it struck a deal with Novell Inc. to get that company's PC networking software to work with IBM mainframes.

DASHED HOPES

Officially, IBM says that the SAA plan is on track. ''I would not categorize SAA as a failure,'' says Joseph Guglielmi, an IBM personal-computer vice-president. He admits that all the ingredients aren't ready yet but says that there are more than enough to get started. Some customers agree: Michael Brown, senior vice-president for information services at The New England, says that the insurer uses the SAA specifications that IBM has published as a ''guideline'' to modernize its systems.

Many buyers had expected more. W. Edward Hodgson, manager of computing and communications at Schindler Elevator Corp. in Morristown, N. J., says he considered buying IBM AS/400 minicomputers but changed his mind because IBM could not deliver the SAA software to make his mainframes work smoothly with the AS/400s. McKesson's d'Alessio passed on the AS/400 for the same reason.

Meanwhile, IBM continues to redefine SAA. In the four years since the plan was put forth, customers have come to demand ways to link all brands of computers -- not just select IBM models. IBM points out that SAA systems can be linked with systems running IBM's version of Unix. And an IBM open-systems group in Austin, Tex., is working on links to other Unix machines. Still, IBM has not officially added Unix -- and most other putative industry standards -- to SAA. ''It's as though they haven't made their minds up what SAA is,'' says Schindler's Hodgson.

What went wrong with SAA? One problem has been sheer size. SAA spelled out rules that gave Wheeler's 30,000 programmers a common goal, but it also created new bureaucracy. Before SAA, IBM programming teams worked independently, not worrying how a mainframe operating system, for instance, fit in with other IBM software. With SAA, however, even IBM's biggest programs became pieces in a giant puzzle. ''I can visualize the enormous number of meetings and technical discussions needed to arrive at what has to be done,'' says Pac Bell's Hancock.

The project has also suffered from IBM's missteps in personal-computer software. Part of the original SAA plan was to adopt a common user interface -- an Apple Macintosh-like graphics format -- to make all SAA applications look the same. That format, called Presentation Manager, was part of the OS/2 operating system, introduced in 1987. But OS/2 didn't catch on, partly because it wasn't compatible with the older MS-DOS software. Now, IBM is pushing a more MS-DOS-compatible update to OS/2. But it still uses Presentation Manager. IBM faces an uphill battle against Microsoft Corp., which is selling MS-DOS with Windows, its popular graphics interface.

In one sense, the SAA problem reflects a larger IBM headache: its inability to move quickly in software and services. In 1986, company executives said that by the mid-1990s, IBM should get half its sales in software and services. In 1990, the figure was 31%, and software revenues were only 14% of sales -- up from 11% in 1986.

UNENDING JOURNEY

Even if IBM can deliver OfficeVision and other critical SAA software within a year or two, that may be too little, too late. Customers are likely to force Big Blue to adopt more open-systems connections, say industry watchers. Those would, in effect, supersede the original SAA plan.

Officially, IBM will never disown SAA, because of the enormous amount of money and energy it has sunk into it -- and because of the customers who are still waiting for it. Indeed, SAA czar Wheeler has called his grand plan ''a journey that will never be complete.'' For customers, that makes waiting for SAA a little like waiting for Godot.