Current DataChangeListeners incur heavy performance overhead, as each change is traversed and analyzed for change. As it turns out, applications usually need to extract only partial change information.

While BUG-2672 is the long-term solution, we need a stop-gap for Lithium, where the application will have control over how granular the analysis is. For this we need to provide a cursor-based API, on top of which the DataChangeListener API will be implemented.