Flexible Message Semantics in Distributed Objects Systems

Research project P 14575 Flexible Message Semantic in Distributed Objects Systems Hanspeter MÖSSENBÖCK 26.6.2000 Current distributed systems (using Java RMI, CORBA, ...) clearly define what it means to send a message from one object to another (message semantics). This may not be flexible enough in nonstandard application areas such as Electronic Business on the Internet. One often needs additional or different message semantics, e.g. asynchronous communication or fault tolerant semantics. In the proposed research project we regard message semantics as first class abstractions that can be composed to define the behavior of a particular message. We do not restrict a system to a fixed set of message semantics (e.g. local, remote, synchronous), but provide full extensibility (e.g. asynchronous, replicated, transactional, logging semantics, etc.). The programmer can either use an existing semantics from a library or can create arbitrary new semantics. By using the Decorator design pattern the user can transparently combine multiple semantics and attach them to a particular message. We even go a step further: the same message can have different semantics on two objects of the same class, and the same message may even lead to the invocation of different methods depending on the client that sends the message. Our Java implementation should offer a visual assembly environment that allows dynamic adaptation of message semantics without having to modify existing Java source code. Message semantics are not defined at compile-time, but the programmer can compose them at runtime using metaprogramming and just-in-time stub generation. This ability allows the system to adapt itself to changing environments without a substantial performance penalty. We plan to produce a distributed objects system that is a potential target for commercialisation: interest in distributed objects is rapidly growing in industry, especially in companies that are committed to internet technologies. The project increases the knowledge related to message semantics in object-oriented systems and provides a test environment for designing new useful semantics. We also expect synergy benefits, because the project combines the technical and practical expertises of the University of Linz and the EPFL Lausanne, with whom we plan to cooperate.