t0, t0. j t0j, messages, which are specied in the sc hema via sigin various contexts. the form = (C; M; P; messg; impl) where Sbehav addition, behavior can be inherited by subclasses, ition is that classes have attached to them a set of and message names can be overloaded, i.e., re-used The second important aspect of an object-oriented So a behavioral schema is a named v e-tuple of database is that it is intended to capture behavior, natures, and which are implemented as methods. In besides structure. To this end, the relevant intuconstructors (like one with an additional array connot specic to object-orien tation. cluding 19; 20; and others. Notice that it [17; 25] mer represent, for example, aggregate information be resolved. Also, implementations typically add a number of additional features, like attributes as database constraints are distinguished). For another instances (a proposal in that direction appears in [3; like an average salary only relevant to the class as a structor to describe matrices), an explicit inclusion needs to be added stating how to conicts should still leaves several aspects open, like single vs. mulexample, the ODMG-93 proposal for a standardized model contains explicit keys, (binary) relation- [10] functions a distinction of class attributes [22; 29], their objects (in particular various forms of composisemantic information on the set of valid databases from instance attributes (the latter are shared by a variety of models proposed in the literature, intiple inheritance; if the latter is desired, a condition This denition resembles what can be found in where object constraints, class constraints, and 4], which every class inherits a distinction between [20], appear in our model, the reason being that these are tion, see integrity constraints which represent [18]), ships, and inverse attributes. None of these features private and public attributes a dieren t set of [12], all objects associated with a class, while the forwhole) a unique root of the class hierarchy from [7], of distinct types of relationships between classes and rently available seem hardly suited for the nonstanpeople had been studying for years already, it seems a single uniform such model, the foundations on for object-oriented databases. Although there is not much of a relational background, and try to exploit dard applications which initiated the consideration and even standardization eorts ha ve recently been personal account of recent work on formal models launched On the other hand, a number of in- [10]. reason seems to be that many researchers have too of object-orientation in the context of databases. A programming-language or knowledge-representation which such models have to be built seem understood, that as long as possible; this is more than conrmed teresting research issues still deserve further investiby the ODMG-93 proposal. As was done a number of years ago, when database people discovered what In this short paper we have tried to give a rough gation. In particular, formal models as they are cur