We introduce a data model for a context-management middleware that enables context-aware and pervasive computing applications to transparently access available data providers and that effectively combines their data. Our approach supports new data fusion concepts for overlapping and heterogeneous data sets and thus maximizes the information presented to the application. The main part of our data model is a flexible concept for meta data that is able to represent important aspects like quality, data derivation, or temporal characteristics of data. Attributes having multiple values are utilized to represent sensor measurements histories like locations of mobile objects at different points in time. In our paper, we characterize the requirements for our data model and show that existing data models, including the (object-) relational data model and standard XML data models, do not offer the required flexibility. Therefore basic XML technology is extended to support the necessary meta data concept and multiply typed objects.