This paper investigates the applicability of User Interaction Diagrams (UIDs) as user scenarios for the speci¯cation of software requirements by non-technical customers. Two methods for building user scenarios using UIDs were proposed: the progressive and the regressive methods. These two methods were applied in an experiment where the results demonstrated that the regressive method requires signi¯cantly less e®ort as compared to the progressive method. Furthermore, there was a signi¯cant di®erence in the quality of diagrams obtained from each of the two methods. In our experiment, the regressive method resulted in better quality factors.

Abstract-This paper investigates the applicability of User Interaction Diagrams (UIDs) as user scenarios for specifying requirements of software built by non-technical customers. User scenarios represent an alternative to representation of Acceptance Test-Driven Development (ATDD). Two methods for building user scenarios using UIDs were proposed: the progressive and the regressive methods. The progressive method for construction of scenarios provides a description from any starting point until the expected result is reached. The regressive method is based on the Assert-First technique, introduced in Test-Driven Development (TDD), where the user scenario is constructed the other way round, that is, from the expected result to the starting point. These two methods were applied in an experiment where the results demonstrated that the regressive method requires significantly less effort as compared to the progressive method. The quality criteria of the two methods were different, where the regressive method yielded better results.

In the software development process, the acceptance testing may be used by non-technician users to define software requirements. In this article, we use the US-UIDs (User Scenarios through User Interaction Diagrams) as automated acceptance tests in order to provide communications and collaboration between programmers and users. We propose three metrics for measuring the data uniformity in the US-UIDs. These metrics are investigated in four projects. The resulting measures from the investigations of the four projects are used to build a scale with classes to classify the uniformity of the US-UIDs. The classes (duplication, uniformity, irregularity) were created from empiric evaluation and compared to the measures from the offered metrics. The classification purpose is to identify the US-UIDs as uniform or irregular.

This paper presents the design and results of an experiment to evaluate the impact/effect of data uniformity in automation of acceptance tests. An experiment to specify acceptance tests, represented by the User Scenarios through User Interaction Diagrams (US-UIDs) format, with non-technical user has been set up involving two projects. In the first project, called P1, the treatment of data uniformity is held by an expert, while in the second project, called P2, no treatment of data uniformity is done. In both projects, automation of acceptance tests was developed for evaluation and comparison of the following artifacts: data uniformity, fixture name sharing, automation time, and glue code volume. The results show that there is a meaningful statistics difference of uniformity between projects P1 and P2, where P1 resulted in a better uniformity. However, although the treatment of data uniformity does not show meaningful statistics difference according to the strategy of fixture names sharing used, the time spent in fixture naming was more than two times higher in P2. In addition, the glue code volume was less than half in P1 comparing to P2.

No abstract