The Harmony system is a generic framework for reconciling disconnected updates to heterogeneous, replicated XML data. It can be used, for instance, to synchronize the bookmark files of several different web browsers, allowing bookmarks and bookmark folders to be added, deleted, edited, and reorganized by different users running different browser applications on disconnected machines.

A central theme of the Harmony project is bringing ideas from programming languages to bear on a set of problems more commonly regarded as belonging to the purview of databases or distributed systems. In particular, the issue of ALIGNMENT during reconciliation---that is, of determining which parts of divergent replicas are intended to represent ``the same information''---can be addressed by focusing on the type structure of the data being reconciled.

This talk discusses Harmony's synchronization algorithm, its specification, and its applicability to synchronization of (trees representing) a variety of specific forms of application data, including sets, records, tuples, and relations.

Further information, live demos, and an open-source implementation can be found on the Harmony home page: http://www.cis.upenn.edu/~harmony.

[Those who were at the Trusted Global Computation meeting at ETAPS '05 will find some of the talk familar, but I've added some new material on synchronizing ordered data such as lists.]