For non-collaborative data sources, both cost estimate-based optimization and quality-driven query processing are difficult to achieve because the sources do not export cost information nor data quality indicators. In this paper, we first propose an expressive query language extension using QML 1 syntax for defining in a flexible way dimensions, metrics of data quality and data source quality. We present a new framework for adaptive query processing on quality-extended query declarations. This processing includes the negotiation of quality contracts between the distributed data sources. The principle is to find dynamically the best trade-off between the cost of the query and the quality of the result retrieved from several distributed sources.