Landry et al. describe a distributed system for scheduling air traffic. The system is greedy, so it does not produce an optimal solution, but it is flexible and operates in near real time, which is important given the application domain. Uncertainty in arrival times, due to weather and air traffic conditions, is the main issue that complicates the task, together with the large number of aircraft that need to be served and the physical constraints on holding patterns.

The algorithm presented is very simple since the distributed scheduler deals with a single aircraft, and synchronization between different aircraft happens implicitly through the sharing of rate profiles. The method has been tested with good results on real field trials and in an operational test, and seems promising for implementation and real-life use.

The paper does not attempt to prove any formal guarantees for the algorithm, such as fairness and avoidance of deadlocks. The application area of aviation usually demands a certain level of formal guarantees. Scalability is mentioned, but has not been studied. The method is specific to the properties of US air traffic control, so it would not be easy to apply it to other scheduling problems.

The paper suffers from minor inconsistencies, making some parts hard to follow. Examples include inconsistencies between the labels in figure 1 and the corresponding text (in the figure there is no sector labeled A), and misdirected arrows in figure 3 (all of the schedulers should send the required times of arrival (RTAs) to the center, not the opposite). In spite of these inconsistencies, the simplicity of the method shines through.