TOWARD A SOFTWARE INFORMATION SYSTEM _______ David G. Belanger, Ronald J. Braehman, Yih-Farn Chen, Prekumar T. Devanbu, and Peter G. Selfridge 22 David G. Belanger, Ronald J. Brachman, Ylh-Farn Chen, Prekumar T. Devanbu, and Peter G. Selfridge are with AT&T Bell Laboratories, Murray Hill, New Jersey. Mr. Belanger is head of the Advanced Software Department. He is responsible for research into concepts and tools to increase productivity and quality in software develop- ment. He received a B.S. from Union Col- lege, an M.S. from Case Institute of Tech- nology, and a Ph.D. from Case-Western Reserve University, all in mathematics. He joined AT&T in 1979. Mr. Brachman is head of the Artificial Intelli- gence Principles Research Department. He is responsible for development of the CLASSIC knowledge representation system, basic research into tractable reasoning systems, and applica- tions of artificial intelli- gence technology to software development. (continued on page 41) AT&T TECHNICAL JOURNAL.MARCHI APRIL 1990 A software information system collects information about large software systems in knowledge and data bases. It includes applications built on these bases to provide easy access to software information or to automate elements ofthe software development process. Work in this area is proceeding at AT&T Bell Laboratories in three related directions. The C Information Abstraction system automatically extracts information from C language pro- grams and stores it in a relational database. It includes a growing collection of applications ofthis information. The MView system provides graphical views of interre- lated software and lets a user quickly browse software systems. The laSSIE system represents large amounts of software information in a formal language and pro- vides intelligent assistance in retrieving software com- ponents. Progress in these three directions is reviewed in this paper. Introduction The growth in the size and complexity of telecommunication software is fast outpacing the ability ofany individual or group of indivi- dualsto understandlarge systems. Many systems contain more than a half million lines of code and range up to large switching systems of 3 to 5 million lines of code each. As the software portion of our products increases, the competitiveness of AT&T productsis increasingly deter- mined by the software that runs them. For example, the 5ESS® elec- tronic switching system, AT&T's largest switch (itcan serve over 100,000 customerlines),is controlled by complex software responsible for providing plain old telephone service (POTS), a variety of additional features, such as call forwarding and data transmission for businesses, IntegratedServices Digital Network (ISDN), and internal functions such as collecting and collating billing information and determining that the switch is functioning correctly.'