This paper addresses the context of enterprise modeling aiming at a methodology showing how a unified model intended for the information integration and interoperability in heterogeneous world can be systematically synthesized. We can observe an explosion of diverse information modeling facilities and respective platforms around the world. Various attempts to reach technical interoperability among the platforms are well known. Diverse middleware frameworks have been developed emphasizing different basic artifacts (objects, messaging, services, etc.). Taking into consideration information modeling semantics in such frameworks is rarely attempted. The approach considered here is based on the following idea. To provide semantic information integration and interoperability in a heterogeneous environment including different modeling facilities an extensible canonical model is required. A core of the canonical model is fixed. Then for each specific information model Mi, taken from the environment, an extension of the canonical model core is built so that this core together with such extension is refined by Mi. Such refinement-based model transformation should be provable. The canonical model for the environment is synthesized as a union of extensions created for all the models Mi in the environment.
Techniques for such approach are considered separately for structured data models with behaviors, for object models and for process models. The paper has a form of a short overview of basic results related to these techniques and obtained at IPI RAS in different periods during more than 20 years. Experience reached, lessons learnt and perspectives of the approach for the Interop project (enterprise modeling) are discussed.