Data from different autonomous systems should be kept consistent to fulfill the business collaboration. This paper presents the framework, called MDSIF, for data synchronization among multi-datasets based on messages. With the shared data center, MDSIF provides the coherent data snapshots for all related datasets by sending, receiving, and storing the shared data encapsulated in messages. Furthermore, it introduces MDSIF's synchronization mechanism in the formal way, which resolves data conflicts by setting attribute confidences and current value confidences respectively. According to this mechanism, those from the dataset with higher confidence value are allowed to overwrite those from the dataset with lower one. Finally, the case based on MDSIF is illustrated, showing that MDSIF needs only minimum manual intervention and is much more flexible, extendable and reliable, compared with the traditional trigger-based approach. © 2009 Springer-Verlag Berlin Heidelberg.