A great deal of work has been carried out in recent years to facilitate access to data and information available on the Web. Proposals converge in two additional areas which consist in providing the sources with semantic annotations and in designing languages and tools that are capable of using these annotations. However, a large number of sources have not yet been annotated suitably. Besides, languages and existing tools do not allow the user to formulate "blind" queries without knowing the sources. To overcome these two limitations, in this paper we propose a flexible query language which allows a user to query sources in an anonymous way without knowing their existence and their structure. Queries can be solved by a system which in advance discovers potential sources and memorizes their schemas. We clarify how such a system can function.