| With the size of application software increasing from a few hundred lines of code to hundreds of thousands of lines of code, software reliability has become a major issue. Poor reliability often causes major or disasterous consequence. One notable example is the failed interception of the incoming Scud by the Patriot missile in the first Gulf War. Although researchers and engineers have devised many software reliability models, the problem of software reliability still remains.This paper takes a different angle in tackling software reliability problem in which software development/execution platforms are taken into consideration. The platforms we consider include instruction set architecture, operating system, compiler, and programming language. We believe that these platforms impose a maximum reliability value on the software developed or run on them. Our ultimate objective is to construct a software reference model that can forecast softawre's maximum achievable reliability value under specific platforms and/or select appropriate platforms to satisfy target reliability objective.This paper focuses on programming language impact on software reliability. In order to clearly define the concept of failure, this paper targets on software's computational reliability, a measure that is closely related to programming language. This paper studies three popular programming languages and compares their impact on software's computational reliability. Our analysis and experiement show that the computational reliability of FORTRAN equals to that of C/C++, while Java's computational reliability is less than C/C++. |