A multidatabase system based on OMG's distributed object management architecture and named Middle East Technical University (METU) INteroperable Database management system (MIND) is presented. MIND employs a dynamic query optimizer. After a query is decomposed, the global subsequeries are sent to the involved Local Database Agent (LDA) objects. The optimization process starts after the first partial result from a LDA object becomes available. For optimizing the intersite operations among these partial results, a statistical inferencing mechanism is used. Intersite operations are performed by Query Processor (QP) objects. There could be as many QP objects running in parallel as necessary.