ABSTRACT

We describe a novel method for semantic-based decomposition of long-lived transactions as a key element of optimistic replication technologies widely applied in advanced collaborative environments. The research presented addresses the general problem of semantically consistent and functionally meaningful reconciliation of divergent replicas and enables significant simplification and formalization of its solution. The advantages of the method are strong guarantees of semantic consistency of the convergent representations, capabilities to use it in autonomous and user interactive modes as well as avoidance of combinatorial explosion peculiar to many other methods. Following hybrid state-operation-transfer approach the method can be successfully implemented within both client-server and stand-alone systems.