Author: Daniel Izquirdo Cortázar, PhD student, at Libresoft, Universidad Rey Juan Carlos, Spain

Abstract: "Little seems to be known about productivity in large, long-lived Free/Libre/Open Source Software (FLOSS) projects. This article presents a preliminary and exploratory study of the relationship between size, on the one hand, and effort, duration and team size, on the other, for 11 FLOSS projects with current size ranging between between 0.6 and 5,3 million lines of code (MLOC). The extracted data does not fit well the cost estimation model COCOMO 81 for proprietary software, motivating the need for FLOSS-specific productivity models. As a first approximation, we evaluated 16 linear regression models involving different pairs of attributes. The major finding of this article is that, if one removes large jumps, a linear model of development (e.g., by expressing effort in contributor-months as a linear function of the number of files) can fit most of the FLOSS projects we analysed. The best effort model we found, had a coefficient of determination R2 = 0.79. One of the more complex FLOSS projects we analysed, namely Eclipse, behaved rather differently than the others, suggesting that different sets of models may be needed for different categories of FLOSS projects. As a by-product of this study, for individual FLOSS projects, we were able to identify growth models with a high degree of accuracy (R2 ≥ 0.98)."