Reconciliation of Shared Data

Nicholas Taylor
University of Pennsylvania

In the Orchestra data sharing system, a large number of peers have local
databases with related content. The peers periodically share the changes
they have made to these local databases with the other peers. Some of
these peers will automatically apply these updates, if they pass a
tuple-level trust predicate supplied by the database operator, and can
be applied to the database without violating constraints. We will
discuss our formalization of the problem, explain some of the issues
that arise when peers modify the same data in different ways, and give a
concrete example of how the system should perform in practice.