Interoperability of Software Documents

The software development process produces a diverse collection of documents ranging from requirements speciications to architecture diagrams to program source code to bug reports. Some are written in formal languages, but others, while highly structured, are written in natural language. The content of these documents is interconnected in complex ways. For instance, a change report might describe how source code was modiied to conform to a design speciication, the problem having been identiied because test output was incorrect. The current state of the art in software development environments uses many diier-ent, and often incompatible, tools to manage these diierent documents. Many signiicant advances in interaction for the software development process will not be possible until all the documents it produces inter-operate. Large software projects produce an impressive array of documents. In a sense, program source code is the most important type of document produced , since it is the most direct expression of the nature of the program itself. But it is often the case that the majority of software documents are not program source code. Table 1 shows a number of examples, categorized by the phases of the \waterfall" model of the software development process 2]. Some of these documents are descriptions of the program (e.g. user documentation and architectural design). Others describe the process