No users no dataspaces! Query-driven dataspace orchestration

Data analysis in rich spaces of heterogeneous data sources is an increasingly common activity. Examples include querying the web of linked data and personal information management. Such analytics on dataspaces is often iterative and dynamic, in an open-ended interaction between discovery and data orchestration. The current state of the art in integration and orchestration in dataspaces is primarily geared towards close-ended analysis, targeting the discovery of stable data mappings or one-time, pay-as-you-go ad hoc data mappings. The perspective here is dataspace-centric. In this paper, we propose a shift to a user-centric perspective on dataspace orchestration. We outline basic conceptual and technical challenges in supporting data analytics which is open-ended and always evolving, as users respond to new discoveries and connections.