Transaction processing is an important topic, not only in the field of databases but also in other areas, such as workflow processing, Web services, and grid computing. This paper--freely available through OpenAccess--provides a historic survey of the transaction models and mechanisms developed over time to support the requirements of different emerging application domains. Specifically, Wang et al. identify five transaction management eras: the Stone Age, Classical History, Middle Ages, Renaissance, and Modern Times.

The Stone Age is characterized by the absence of transaction models and mechanisms, so it is not explicitly studied in the paper. In the Classical History era, the focus is on transaction management in databases, motivated by the need to ensure the reliability of data-centric applications in multi-user environments. The Middle Ages implies the appearance of advanced transaction models that are more appropriate to perform long and complex tasks; these new models divide a transaction into subtransactions (component transactions), so that in case of failure, the system can avoid restarting the whole transaction from scratch. The Renaissance is marked by the development of workflow management systems--business process modeling and execution. Finally, the Modern Times era is more Internet oriented and includes Web services transactions and grid transactions. Besides the historic review of transaction models, the paper also includes a description of some transaction frameworks “that encompass more than a single transaction model,” for example the Java 2 Enterprise Edition (J2EE) transaction framework.

The paper is well written and easy to read, providing a good starting point for researchers, as well as students interested in transaction management or any of its application fields. However, the descriptions and explanations are sometimes brief--due to the unavoidable space constraints--so significant background knowledge is required to fully understand the paper. In this regard, I found the conclusions presented at the end of the analysis of each era, which make the paper much more accessible to readers who decide not to study the presented approaches in detail, particularly interesting. The authors of this survey work actively in the area of transaction management for business processes. Although they do not aim to provide an exhaustive analysis of all existing approaches for transaction management--for example, the reader interested in mobile transactions is referred to another survey--they fulfill their goal of providing an interesting historical overview.