AbstractSome recent papers have included eulogies of the PDP‐11 architecture, and of its interrupt mechanism in particular. This is dangerous if it is to set a standard for the coming decade. This note presents the view that all interrupts should be serviced by dedicated processes. However, the feasibility of this is determined by machine architectures, not just by the hardware interrupt mechanism itself.Considering first the well‐known PDP‐11, we then summarize the origin and function of interrupts before stating a philosophy for interrupt‐handling. It is then shown, with reference to the PDP‐11, how not only the interrupt mechanism but also features such as stacks and procedure mechanisms may affect this philosophy. We conclude by suggesting alternative architectural features to unify interrupts, procedure calls, process invocations, and interprocess communication, as will be required for the support of recent and future languages.