Abstract The amount of code in evolving software‐intensive systems appears to be growing relentlessly, affecting products and entire businesses. Objective figures quantifying the software code growth rate bounds in systems over a large time scale can be used as a reliable predictive basis for the size of software assets. We analyze a reference base of over 404 million lines of open source and closed software systems to provide accurate bounds on source code growth rates. We find that software source code in systems doubles about every 42 months on average, corresponding to a median compound annual growth rate of 1.21 ± 0.01. Software product and development managers can use our findings to bound estimates, to assess the trustworthiness of road maps, to recognise unsustainable growth, to judge the health of a software development project, and to predict a system's hardware footprint.

In this paper it is proposed that the Conservation of Hartley-Shannon Information (hereafter contracted to H-S Information) plays the same role in discrete systems as the Conservation of Energy does in physical systems. In particular, using a variational approach, it is shown that the symmetry of scale-invariance, power-laws and the Conservation of H-S Information are intimately related and lead to the prediction that the component sizes of any software system assembled from components made from discrete tokens always asymptote to a scale-free power-law distribution in theuniquealphabet of tokens used to construct each component. This is then validated to a very high degree of significance on some 100 million lines of software in seven different programming languages independently of how the software was produced, what it does, who produced it or what stage of maturity it has reached. A further implication of the theory presented here is that the average size of components depends only on their unique alphabet, independently of the package they appear in. This too is demonstrated on the main data set and also on 24 additional Fortran 90 packages.

This paper begins by modeling general software systems using concepts from statistical mechanics which provide a framework for linking microscopic and macroscopic features of any complex system. This analysis provides a way of linking two features of particular interest in software systems: first the microscopic distribution of defects within components and second the macroscopic distribution of component sizes in a typical system. The former has been studied extensively, but the latter much less so. This paper shows that subject to an external constraint that the total number of defects is fixed in an equilibrium system, commonly used defect models for individual components directly imply that the distribution of component sizes in such a system will obey a power-law Pareto distribution. The paper continues by analyzing a large number of mature systems of different total sizes, different implementation languages, and very different application areas, and demonstrates that the component sizes do indeed appear to obey the predicted power-law distribution. Some possible implications of this are explored.

This paper describes some results of what, to the authors' knowledge, is the largest N-version programming experiment ever performed. The object of this ongoing four-year study is to attempt to determine just how consistent the results of scientific computation really are, and, from this, to estimate accuracy. The experiment is being carried out in a branch of the earth sciences known as seismic data processing, where 15 or so independently developed large commercial packages that implement mathematical algorithms from the same or similar published specifications in the same programming language (Fortran) have been developed over the last 20 years. The results of processing the same input dataset, using the same user-specified parameters, for nine of these packages is reported in this paper. Finally, feedback of obvious flaws was attempted to reduce the overall disagreement. The results are deeply disturbing. Whereas scientists like to think that their code is accurate to the precision of the arithmetic used, in this study, numerical disagreement grows at around the rate of 1% in average absolute difference per 4000 fines of implemented code, and, even worse, the nature of the disagreement is nonrandom. Furthermore, the seismic data processing industry has better than average quality standards for its software development with both identifiable quality assurance functions and substantial test datasets.>

Abstract The portability of software has become a major commercial issue in recent times. Such portability does not come easily, as few if any computer languages are really portable in practice. An additional complicating factor, especially in the commercial environment, is that the resulting software must be efficient. This paper describes the design and successful implementation of a 500,000+ line portable FORTRAN 77 package for the processing of seismic data. The package exhibits demonstrably high efficiency on a wide variety of machines from minicomputers to the largest supercomputers. Experiences gained during this exercise throw much light on the integration of the various thought processes which occur during the software engineering cycle, especially the notion of locality.