Conversión COSMIC/IFPUG

Abstract

Measurement is a key factor in order to give to Software Engineering the solid, systematic, disciplined, and quantifiable base that software development, operation and maintenance requires. One of the oldest software measures is the functional size measure. The functional size is the size of the software derived by quantifying the functional user requirements. The first method available to measure the software functional size, called function points analysis, was proposed by Allan Albrecht and his collaborators, late in the seventies. With minor changes, the method has evolved up to nowadays and has relatively high diffusion and acceptance. IFPUG - International Function Points Users Group- is the organization actually responsible for the standardization and promotion of the method. A new method for functional size measurement, called full function points, was proposed in recent times by Charles Symons, Alain Abran and others late in the nineties. An improved version of this method is actually standardized and promoted by COSMIC -Common Software Measurement Consortium-. One of the most common uses of functional size measures in software community is for estimating project schedules, required resources, budgeting, and so on. To enable these estimations, there are large databases containing information about finished projects, measured with the IFPUG method. Nevertheless, there are almost no projects measured with the COSMIC method, which discourages usage by community, in spite of the attractiveness the method could have. This work proposes a mechanism to convert the measures of applications performed with the IFPUG method to the measures that would result from measuring the same applications with COSMIC. Supporting such mechanism we perform an exhaustive analysis and comparison of both methods. The analysis has a qualitative component, where we examine and maps concepts between methods, and a quantitative component, where we perform statistics analysis of a relatively large set of applications measured with both methods. The conversion model resulting from this analysis is further validated experimentally.