And this 85 percent of the software industry bears a strong resemblance to many industries of the early 19th century -- virtually all of the products are expensive, hand-crafted, error prone, one of a kind. Large industrial software systems tend to be monolithic. Individual pieces of the system cannot easily be reused (except perhaps in the same company) because of the close cooperation required between the programmer and the systems integrator. By most estimates, over half of all application development projects in this industry end in failure -- after all the time and money is spent, the product still cannot be used operationally.

The goal of the ATP Component-Based Software program is to develop the technologies needed to enable systematically reusable software components -- relatively small, carefully engineered software elements suitable for a broad array of applications. These technologies would enable software companies to build specialized components that can be sold to systems integrators and custom builders, who would combine them with other, largely purchased, off-the-shelf components to create high-quality custom applications.

From the point of view of system integrators and custom end users, software tools would automatically match components to applications systems, ensuring compatibility and reliability. Numerous sources of cost and error in application development would be eliminated.

Major applications would no longer be large monolithic structures built from the ground up but rather assemblies of smaller components, competitively purchased from vendors who would have the specialized skills to concentrate on issues such as quality and reliability for the components they provide.

Far from relying on today's profusion of software interface standards -- which many times have life expectancies shorter than the standards-setting process -- the proposed technology should make many existing software interface standards unnecessary.

Initial goals for the Component-Based Software program -- aspects that would be enabled in the next five to seven years -- include:

• enabling automated assembly of software components, bringing greater quality and reliability, reduced time to develop and test new systems, and cost amortization from multiple use of software components, which would lead to increased productivity in software development;

• enabling increased productivity for software users, brought about by increased quality of components, increased dependability of systems; and

• enabling systematic reuse of components, increased interoperation of software, and easier adaptation of software for foreign markets, which may lead to expanded markets for U.S. software products.

In the longer run -- maybe 10 years after successful completion of the ATP program -- an established software-components industry would provide a rapid, responsive channel for marketing software innovations, while constantly improving quality, reliability, and capability, in much the same fashion that the semiconductor component industry has constantly reduced cost and improved performance of computer chips since the invention of the integrated circuit, through rapid infusion of technical innovations.

Examples from the physical world often are invoked to explain software systems -- in some systems, for example, standard software interfaces for network communications are called sockets. The analogies are insidious, however, because the realm of software is nothing like the realm of physical objects. The idea of manufacturing only one kind of light bulb, which would be reconfigured automatically at the point of use to fit into any desired socket at any voltage, is absurd. But for software components, the idea is not only feasible -- it could be the necessary enabling technology for the complex systems software needed for new, advanced information technology applications.

The ATP program in Component-Based Software will enable this sort of protean software by combining formal semantic methods, which are effective but difficult to use, with more innovative automated methods for software transformation and synthesis that would handle the actual composition of software systems from formal specifications of the desired product and the available components.

The technical goals of the Component-Based Software program include:

• The Component-Based Software program addresses problems -- including quality, cost, interoperation, labor-intensity, and the lack of a viable software-components industry -- which are pervasive throughout all large-scale software developments, but it involves broad-based solutions and significant technical risks that are beyond the scope of any one company or project.

• Any technology that addresses these pervasive problems will succeed only if it is used broadly throughout the industry, but this requires a program that can gain the support and cooperation of competing companies to overcome non-technical barriers to its use. Through the Component-Based Software program, the ATP creates a needed forum for many individual software firms to cooperate on problems.

• Software is a key element of the technology infrastructure, so ATP funds in this area are highly leveraged. The Component-Based Software program, if successful, will enable important advances in many other economically significant areas, including manufacturing systems, healthcare information systems, and the National Information Infrastructure