Distributed computer control systems (DCCS) have a number of potential advantages over centralised systems, especially where the application is itself physically distributed. A computer station can be placed close to the plant being controlled, and a communications network used to enable the stations to communicate to co-ordinate their actions. However, the software must be carefully designed to exploit the potential advantages of distribution. In the paper, the CONIC architecture for DCCS is described, concentrating on the software structure but also briefly describing the physical architecture designed to support a CONIC system. The software structure emphasises the distinction between the writing of individual software components and the construction and configuration of a system from a set of components. A modular structure is used to separate programming from configuration. Typed entry and exit ports clearly define a module interface which, like the plugs and sockets of hardware components, permit modules to be interconnected in different ways. On-line modification and extension of the system is supported by permitting the dynamic creation and interconnection of modules. Message-passing primitives are provided to permit modules to co-ordinate and synchronise control actions.