This paper studies the various aspects of query optimization in deductive object-oriented databases. First, a rule language is proposed for object- oriented databases. This language is a pragmatic C++ adaptation of previous proposals. The supported daw model is the class model of C+ +. The language supports recursive rules, method calls and non monotonic updates. Given a query, an optimizer must compile a rule program into an object algebra program. Tran:-,jormation techniques can then be applied to generare equivalenc programs. Possible transformations include operation permutation, Jixpoint reduction, integrity constraint addition, predicare simplification and method calt optimization. In case of recursive updates, additional memory-based nets are useful to accelerate query eva/uation. Possible choices are Rete, Treat or Cosma networks. We discuss and compare these types of accelerators. The search strategy component is in charge of determining the best choice in the solution space of ali evaluation approaches and access plans. Finally, we conclude by discussing engineering techniques to develop an imegrated optimizer.