8 Conclusions and Future Work

We think that the approach to safe system extension proposed in this paper presents many advantages. From a software engineering perspective, the architectural principles presented in Section 3 considerably simplify the task of building and extending theorem provers safely. The concept of OMRS is a rst small step towards complex reasoning systems built by composing existing modules, or extended with new modules in a \plug and play" (open) manner. The lifting-reasoning-attening cycle is a very general methodology towards the construction of self-modifying systems. This methodology allows us to concentrate on selected parts of the code at the chosen level of detail. Future work will be devoted to the reenement of the ideas presented in this paper. One line of research will aim at completing our characterization of the control and interaction components of a prover, and of the corresponding levels in the OMRS setting. We plan to extend the work on MT in order to express provisional and schematic reasoning, mixing of backward and forward reasoning. In the long run we aim at applying our methodology for safe system extension to parts of a generic theorem prover diierent from the control component. One example is the interaction component; preliminary to this step is however the development of the control level of OMRS. This will provide us with the reference model for the design of the interaction functionalities.