Software Engineering Research Group Kassel

Building a rich internet application (RIA) requires the programming of various callbacks and listeners. AJAX like server requests require callback handler objects that react to the asynchronous server response. Active GUI elements like buttons or menu entries require action handlers. Using a timer queue requires appropriate event handlers, too. Programming all these handlers is tedious and error prone. Subsequent steps of e.g. initialization or of a protocol of server requests are scattered over multiple separated blocks of code. The control flow between these blocks is hard to retrieve. Some common variables have to be introduced in order to pass the application state between the different handler blocks. To overcome these problems, we propose to use an extension of UML statecharts, called action charts, dedicated to the modeling of callbacks and listeners. All kinds of handlers are modeled in a common uniform statechart notation. States become actions or handlers. Transition represent the flow of execution. Variables are shared between actions providing a simple mechanism for passing the application state from one handler to the next. From such action charts we generate GWT compliant Java code that facilitates validation and debugging of the modeled behavior. This is implemented as part of the open source CASE tool Fujaba.

Originally, model versioning has been developed to enable teams of developers to work on common model data, concurrently. We have the idea to use the same techniques to facilitate the collaboration of collaboration applications. Multi threaded applications share a common main memory. Thus, all threads have access to the full data structures and each thread may query and update the data structures, concurrently, in order to fulfill its tasks. In distributed applications, each distributed process has access only to its own share of the data model. In order to query and update remote data structure parts, the process has to send an appropriate request to the process, that owns that data. Transferring complex data structures, e.g. as query result, from one process to the other requires tedious data serialization and deserialization mechanisms. To overcome these problems, this paper proposes to replicate model data for each process and to use model versioning techniques to synchronize the different model data replicas. We have built a web based workflow editor and a web based version of a Ludo game to validate this idea. This paper reports about our experiences with the data replication approach and our experiences in using it for web applications.

The latest research activities of the Fujaba group of Kassel University led to challenges in development of new web tech- nologies enabling end users to wrap services into web gad- gets and to combine them into complex web applications. Web applications running inside a webbrowser bring new requirements to the traditional desktop application devel- opment process. Since web applications usually don’t come to life using model based approaches or Story Driven Mod- elling the Fujaba Toolsuite has to be adopted to fulfil the new requirements. The Fujaba group of Kassel wanted to get all the help we are used to get from Fujaba in ordinary application develop- ment for our web applications, too. Thus, we are developing a new code generation for the generation of Google Web Toolkit compliant Java code that will then be compiled into JavaScript running on web browsers. In addition, we develop tool support for building the GUIs of such web applications. And, on the server side, we develop technologies to facilitate service development with Fujaba. This paper reports on the design of these new Fujaba web components.