This seminar was based on and continued the interaction of different computer-science communities that was begun in an earlier Dagstuhl seminar in April 2004. Both seminars have aimed at a deeper understanding of the fundamental concept of atomic actions and their roles in system design, execution, modeling, and correctness reasoning, and at fostering collaboration, synergies, and a unified perspective across largely separated research communities. Each of the two seminar brought together about 30 researchers and industrial practitioners from the four areas of database and transaction processing systems, fault tolerance and dependable systems, formal methods, and to smaller extent, hardware architecture and programming languages. The interpretations and roles of the atomicity concept(s) vary substantially across these communities. For example, the emphasis in database systems is on algorithms and implementation techniques for atomic transactions, whereas in dependable systems and formal methods atomicity is viewed as an intentionally imposed or postulated property of system components to simplify designs and increase dependability. Nevertheless, all four communities share the hope that it will eventually be possible to unify the different scientific viewpoints into more coherent foundations, system development principles, design methodologies, and usage guidelines.

The 2004 seminar was very successful on connecting the communities. It raised first skepticism and then curiosity about each other's viewpoints and methodologies. As a major achievement, it led to a strategic position paper, entitled "The Atomic Manifesto: a Story in Four Quarks", which appeared, in identical form, in widely read publication venues in the different communities: ACM SIGMOD Record, ACM Operating Systems Review, the Journal of Universal Computer Science, and also within the Dagstuhl Seminar Proceedings. In addition, the seminar produced a special issue of the Journal of Universal Computer Science with 8 full papers that were presented in preliminary form at the seminar and one full paper co-authored by two researchers who had not met before the seminar. The Atomicity seminar in March 2006 was intended to intensify and extend this barely blooming plant of cross-community collaboration, to revisit and refine the technical challenges identified in 2004, and to discuss the progress made in the last two years.

Results of the Seminar

The presentations and discussions at the seminar reconfirmed that a unified foundation for atomicity is a strategically important and pressing research avenue. Furthermore, the seminar was again successful in spawning new collaborations among participants, some of which span communities. It is planned to prepare another special issue for the Journal of Universal Computer Science, with full papers that hopefully emerge from this ongoing work and the results presented at the seminar.

In terms of specific research issues, the following topics led to intensive discussions and were identified as key directions within the broad theme of atomicity:

1. the integration of open nested transactions into programming languages and their run-time environment,
2. methods for providing strong guarantees about system behavior based on weaker guarantees by the underlying components,
3. handling atomic actions in time-critical environments like operating system kernels.