In this paper we describe our initial work on a long-term project to develop and validate a reliability model and a new class of software complexity metrics which are related to this model. In contrast to previous "black box" approaches, the reliability model is novel because it incorporates knowledge about the system in the form of quantitative software complexity metrics. While the initial model uses existing software metrics a parallel effort in this project is investigating new classes of metrics, interface and dynamic metrics, which are useful in their own right but are also of particular relevance to the reliability model. The initial definitions of both the model and the metrics are given along with a description of the next research milestones.