Towards Testing and Analysis of Systems that Use Serialization

Abstract

Object serialization facilitates the flattening of structured objects into byte streams and is therefore important for all component-based applications that strongly rely on data-exchange among components. Unfortunately, implementing and controlling the serialization mechanisms may expose the software to subtle faults. This paper paves the way towards testing and analysis techniques specifically tailored to the assessment of software that uses serialization. In particular, we introduce a taxonomy of abstractions and terms to semantically characterize and classify the main data-exchange cases, which serialization can be involved with. The resulting conceptual framework provides a means to forecast how erroneous implementations of serialization would look like in different cases, thus enabling the focusing of testing and analysis techniques to address serialization-related faults. © 2004 Elsevier B.V. All rights reserved.