Project initially started by Alin Deutsch, Lucian Popa, Arnaud Sahuguet and Val Tannen.

Maintained by Alin Deutsch and Nicola Onose.

MARS is a system for publishing as XML data from mixed (relational+XML) proprietary storage, while supporting redundancy in storage for tuning purposes. The correspondence between public and proprietary schemas is given by a combination of LAV and GAV-style views expressed in XQuery. XML and relational integrity constraints are also taken into consideration. Starting with client XQueries formulated against the public schema, MARS achieves the combined effect of rewriting-with-views, composition-with-views and query minimization under integrity constraints to obtain optimal reformulations against the proprietary schema.
MARS achieves the uniform treatment of XML and relational storage by first reducing the reformulation problem to that of minimization of relational queries in the presence of relational constraints. The relational minimization is next performed by a core module which implements the Chase&Backchase algorithm.

We are currently publishing the Chase&Backchase core module. We plan to release the reduction module in the near future.
In the interim, the current MARS release can be used as is for reformulation in relational database publishing. For XML publishing, one must encode by hand the XQuery views and the XML constraints into relational constraints.