Creating and Using Models for Mobile and Rich Internet Applications

点击翻译按钮获取中文摘要

The so-called "Rich Internet Applications" (RIAs) are increasingly used for modern applications. They have the ability to update the client-side programmatically and to interact with the server asynchronously. In effect, RIAs are client-server applications, and with wider adaption of HTML5, RIAs will be increasingly stand-alone. Automatically inferring a model (a.k.a. "crawling") for these RIAs is necessary for tasks such as content indexing, security testing, model-based testing etc. The bulk of the recent research in this domain is based on Finite State Machines (FSM)-based models, where each state represents one possible "Document Object Model" (DOM) of the application, and each transition represents the execution of one JavaScript event. However, these models are unable to scale up to the size of most real RIAs. In this research, we want to improve model-inference techniques for RIAs and Mobile applications to be able to handle large, complex RIAs and then use these models in the context of automated formal application testing. The first axis is to build upon our new model, which we have called "Component-Based". The goal is to automatically break down each DOM into a set of independent components, which are then modelled independently from one another. This dramatically reduces the size of the model when compared to the DOM-based model, since it avoids much of the unnecessary duplication inherent to these models. Our initial evaluations have shown that most modern complex RIAs have many of these components, making this model highly applicable. We will in particular work on formally defining this new model and its properties. We will also focus on automatic component detection and efficient modelling algorithms. Our second axis is automatic user-session reconstructions based on execution logs. This provides the ability to recover the actions of user's offline using common log files. This presents several benefits, including the production of better models (input values, possible sequences of actions etc.), the automatic production of necessary information for further automation (e.g. the ability to infer complete login sequences for automatic testing) and automatic reproduction of faults for debugging purposes. Our last research axis is to use these models in the context of formal Model-Based Testing. We are particularly interested in the generation of checking sequences in this context (a sequence that proves the conformance of the implementation to our model, under a given fault model).  The results of this research will be of interest to researchers and practitioners alike. It will allow us to expend on an already successful research program, giving us the opportunity to work on topics of general interest in this domain while continuing our shorter-term work with IBM in parallel. Canada is a research leader in the domain of RIAs, and this research program will strengthen that position.