On the quality of Web Services.

Web Services (WSs) are gaining increasing attention as programming components and so is their quality. WSs offer many benefits, like assured interoperability, and reusability. Conversely, they introduce a number of challenges as far as their quality is concerned, seen from the perspectives of two different stakeholders: (1) the developer/provider of WSs and (2) the consumer of WSs. Developers are usually concerned about the correctness of the WS's functionality which can be assessed by functional testing. Consumers of WSs are usually careful about the reliability of WSs they are depending on (in addition to other qualities). They need to know whether the WSs are available (i.e., up and running), accessible (i.e., they actually accept requests) while available and whether they successfully deliver responses for the incoming requests. Availability, Accessibility, and Successability of WSs are directly related to WS reliability. Assessing these three factors via testing is usually only feasible at late stages of the development life-cycle. If they can be predicted early during the development, they can provide valuable information that may positively influence the engineering of WSs with regards to their quality. In this thesis we focus on assessing the quality of WSs via testing and via prediction. Testing of WSs is addressed by an extensive systematic literature review that focuses on a special type of WSs, the semantic WSs. The main objective of the review is to capture the current state of the art of functional testing of semantic WSs and to identify possible approaches for deriving functional test cases from their requirement specifications. The review follows a predefined procedure that involves automatically searching 5 well-known digital libraries. After applying the selection criteria to the search results, a total of 34 studies were identified as relevant. Required information was extracted from the studies, synthesized and summarized. The results of the systematic literature review showed that it is possible to derive test cases from requirement specifications of semantic WSs based on the different testing approaches identified in the primary studies. In more than half of the identified approaches, test cases are derived from transformed specification models. Petri Nets (and its derivatives) is the mostly used transformation. To derive test cases, different techniques are applied to the specification models. Model checking is largely used for this purpose. Prediction of Availability, Accessibility, and Successability is addressed by a correlational study in which we focused on identifying possible relations between the quality attributes Availability, Accessibility, and Successability and other internal quality measures (e.g., cyclomatic complexity) that may allow building statistically significant predictive models for the three attributes. A total of 34 students interacted freely with 20 pre-selected WSs while internal and external quality measures are collected using a data collection framework designed and implemented specially for this purpose. The collected data are then analyzed using different statistical approaches. The correlational study conducted confirmed that it is possible to build statistically significant predictive models for Accessibility and Successability. A very large number of significant models was built using two different approaches, namely the binary logistic regression and the ordinal logistic regression. Many significant predictive models were selected out of the identified models based on special criteria that take into consideration the predictive power and the stability of the models. The selected models are validated using the bootstrap validation technique. The result of validation showed that only two models out of the selected models are well calibrated and expected to maintain their predictive power when applied to a future dataset. These two models are for predicting Accessibility based on the number of weighted methods (WM) and the number of lines of code (LOC) respectively. The approach and the findings presented in this work for building accurate predictive models for the WSs qualities Availability, Accessibility, and Successability may offer researchers and practitioners an opportunity to examine and build similar predictive models for other WSs qualities, thus allowing for early prediction of the targeted qualities and hence early adjustments during the development to satisfy any requirements imposed on the WSs with regards to the predicted qualities. Early prediction of WSs qualities may help leverage trust on the WSs and reduces development costs, hence increases their adoption.