Data dependencies (or constraints) in the relational model are the basis of the database conception theory and their practical consequences are remarkable. The most famous dependencies are functional dependencies (FD), inclusion dependencies (ID) and multivalued dependencies. Many more general classes were proposed like tuple generating dependencies (TGD) in order to extend their expressiveness in a unified way. Theses dependencies aim to capture some constraints that are to be verified by conventional data.

With the expansion of the Web, the democratization of producing and accessing data, applications are no longer using data only stored on hard drive disk. They can require non classical data like streams (e.g. from sensors) and services available in their environment (e.g. sending messages to somebody using instant messaging). In this context, conventional dependencies are not anymore sufficient to describe the constraints of these new applications.

Within the SoCQ project at LIRIS, we proposed an extension to relational algebra allowing the manipulation of data, streams and services in a unified fashion. Such algebra as an impact on query languages but also on the conception of the data structure. Our work within SoCQ project is to extend classical dependencies for the modeling of applications dealing not only with data, but also with streams and services.