Annals of Software Engineering 14, 93–114, 2002 2002 Kluwer Academic Publishers. Manufactured in The Netherlands.Testing Processes of Web ApplicationsFILIPPO RICCA and PAOLO TONELLA {ricca, tonella}@itc.itITC-irst, Istituto per la Ricerca Scientifica e Tecnologica, Via alla cascata, 38050 Povo (Trento), ItalyAbstract. Current practice in Web application development is based on the skills of the individual program-mers and often does not apply the principles of software engineering. The increasing economic relevanceand internal complexity of the new generation of Web applications require that proper quality standardsare reached and that development is kept under control. It is therefore likely that the formalization of theprocess followed while developing these applications will be one of the major research topics.In this paper we focus on Web application testing, a crucial phase when quality and reliability area goal. Testing is considered in the wider context of the whole development process, for which an incre-mental/iterative model is devised. The processes behind the testing activities are analyzed considering thespecificity of Web applications, for which the availability of a reference model is shown to be particularlyimportant. The approach proposed in this paper covers the integration testing phase, which can take advan-tage of some features of Web applications (e.g., the http protocol employed), thus resulting in a higher levelof automation with respect to traditional software.The testing processes described in this paper are supported by the prototype research tool TestWeb. Thistool exploits a reverse engineered UML (Unified Modeling Language) model of the Web application togenerate and execute test cases, in order to satisfy the testing criteria selected by the user. The usage of thistool will be presented with reference to a real-world case study.1. IntroductionMost Web applications have insofar been developed without following a formalizedprocess model [Pressman 2000]. Requirements are not captured and the architecture anddetailed design of the system are not considered. Developers quickly move to the imple-mentation phase and deliver the system without testing it. No documentation is usuallyproduced about the internal organization of the application. While this kind of practicewas motivated by the characteristics of the first generation of Web sites, now things arechanging and there is an increasing demand for better techniques, methodologies andprocesses.The delivery of Web based systems developed according to ad hoc methods andwith no consideration to the principles of software engineering was justified by the sizeof these applications, which were typically small, by the estimated lifetime, which wasexpected to be short, and by the difficulties to capture the user needs, both those to besatisfied by the initial release of the application and those that were considered likely toemerge in the future. Moreover, the first generation of Web sites were little more thanfixed advertising material made publicly available: they had no particular relevance forthe core business.94 RICCA AND TONELLASoon companies realized that the Web is not just a way to promote their image,but can be exploited as a means to provide services. At the same time, several tech-nologies have been developed to support the production of increasingly complex Webapplications, which can be effectively exploited to support the main company business.Consequently, these applications have begun to be critical for the companies and to in-corporate advanced functionalities. The next step in this evolution path will be to absorbsome of the lessons learned during the history of software engineering, which startedbeing considered when the practice of software development suffered similar problemsas the current one for Web based systems.While substantial effort was devoted to investigating models and formalisms aimedat supporting the design of Web applications [Bichler and Nusser 1996; Conallen 2000;Isakowitz et al. 1997], only few works considered the problems related to Web sitemaintenance [Antoniol et al. 2000; Warren et al. 1999] and testing [Chang and Hon2000; Liu et al. 2000; MacIntosh and Strigel 2000; Miller 1998].The existence of problems, in Web site development, similar to those encounteredin software before the advent of software engineering are recognized in [Warren et al.1999], where the evolution of Web sites is characterized by means of metrics.Statistical testing is proposed in [Chang and Hon 2000] for the automatic selectionof the paths to be exercised in a Web application. The number of invalid links (discussedin section 4) encountered along the test paths allows estimating the site reliability, i.e.,probability that a user completes the navigation without errors. The CAPBAK/Web tool,explained in [Miller 1998], is a Web testing tool that supports functional testing and re-gression testing. In [Liu et al. 2000] an approach to data flow testing of Web applicationsis presented. In their approach, the structural test artifacts of a Web application are cap-tured in a Web application Test Model where each component of a Web application isconsidered as an object. Data flow test cases for Web applications are derived from flowgraphs in five levels according to the types of definition use chains for the variables ofinterest.In this paper the testing processes of Web applications are considered in the largercontext of Web application development, for which the availability of a reference modelis central to several activities, such as understanding of the existing system, assessmentof the required changes and implementation of the modifications. The testing phase isalso expected to benefit from a structural model of the application, which is the startingpoint for white-box testing. In particular, we will focus on integration testing of Webapplications, where the specific features of these applications, compared with traditionalsoftware, can be exploited to improve the level of automation and simplify the testingprocedures. A support tool will be presented with reference to a real world case study,showing that a practical use of the proposed techniques is affordable.The paper is organized as follows. Section 2 provides a candidate process model forthe development of Web applications, based on their specific characteristics. Section 3describes the reference model that will be used for testing. An