Panel ReportThe panel opened with a question from the audience asking how the evolvability of Autonomic Communications (AC) and the stability of the resulting architectures and systems can be ensured. The panel responded stating that stability can not be regarded in terms of deterministic system configuration, but needs to be viewed in terms of behavioural stability. Thus we must tolerate a level of volatility but only within a well understood behavioural envelope that relates to specific autonomic tasks. In other words, we should focus on enforcing specific bounds on the adaptivity that selfmanaged systems may exhibit, rather than on achieving full behavioural determinism. With respect to the evolution toward and evolvability of AC, it was agreed that gradual changes were a real-life necessity. As a result we require ways to subdivide AC architectures into separate areas of concern that can be attacked, solved and deployed independently. However, there were no immediate suggestions for the lines along which such a separation would best be made. Preceding the panel, a poster presentation had included a synthesis of issues raised during the workshop in the form of a layered cube reminiscent of that used to explain broadband ISDN principles during the 1990's. It was observed that this synthesis served to show the potential complexity and inter-connectiveness of issues in AC. Reactions to this model, however, also hinted at the challenges in defining any clear architectural separations for AC given our current understanding of the field. It also spurred comments on the lessons that could be learnt by the failure of ATM to reach its technical potential due to a lack of flexibly in reacting to changing economic and market concerns. There was