Correspondence and translation for heterogeneous data

We presented a specification of the integration of heterogeneous data based on correspondence rules. We showed how a unique specification can served many purposes (including two-way translation) assuming some reasonable restrictions. We claim that the framework and restrictions are acceptable in practice, and in particular one can show that all the document-OODB correspondences/translations of [2, 3] are covered. We are currently working on further substantiating this by more experimentation.When applying the work presented here a number of issues arise such as the specification of default values when some information is missing in the translation. A more complex one is the introduction of some simple constraints in the model, e.g., keys.Another important implementation issue is to choose between keeping one of the representations virtual vs. materializing both. In particular, it is conceivable to apply in this larger setting the optimization techniques developed in a OODB/SGML context for queries [2] and updates [3].