A lot of discussion and debate exists in embedded software development about the relative merits of two interrupt handling architectures. The debate surrounds the “unified” and “segmented” models, which take very different approaches to how the critical sections of a real-time operating system are protected during interrupt servicing. This paper examines these two interrupt handling models, and introduces a novel hybrid that combines the best features of both. This hybrid does not require any changes to the user application and operates automatically with zero overhead.