To design a conceptual schema, that is, to establish and describe the necessary propositions of the universe of discourse, the information system designer starts by observing the selected portion of the world constituting the universe of discourse and constructing in his mind an abstraction of it, much in the same way as a scientist, by experiment and analysis, constructs a theory of the observed phenomena. As in science, we would normally expect that such an abstraction would be general enough to encompass many different (preferably all) occurrences of the same kind of phenomena. This abstraction constitutes propositions that necessarily hold in all possible entity worlds, just as the theory must hold for the observed or expected phenomena. The process of observation, abstraction, and conceptual schema formulation is usually iterative. This is similar to the scientist's attitude during the formulation of a theory: further experimentation - that is, observation of the universe of discourse - is generally required to clarify some aspects of the theory. Although selection of what is considered to be necessary propositions about the universe of discourse to be described in the conceptual schema is to a certain extent arbitrary, as already is mentioned in chapter 1, the systems designer might consider various factors in deciding the boundary of the the conceptual schema. An example of a widely accepted and also basic distinction has already been mentioned in chapter 1. It is the distinction between sentences which represent general laws and rules to which possible entities in the universe of discourse have to adhere, and sentences which describe facts about particular entities in the universe of discourse, following the laws and rules described in the conceptual schema. The main reasons for this distinction are practical arguments with respect to the design, implementation and maintenance of the information system, because some sentences are generalizations (conceptual schema) and others are specifics (information base). These generalizations have already been listed in chapter 1, section 1.8, as: - describing classes (types, variables) in the universe of discourse rather than individuals (instances), - describing concepts that are less subject to change rather than concepts that are changing more frequently, - inclusion of rules or constraints having wide influence on the behaviour of the universe of discourse (and therefore on the behaviour of the conceptual schema and information base) rather than narrow influence. The guiding principle in selecting the necessary propositions is that they, although stating something about the universe of discourse, are convenient in controlling the consistency of the collection of sentences in the information base and the permissibility of their manipulation. As already stated in section 1.8 of chapter 1, the following general principles for the conceptual schema should be observed at all times: * 100 Percent principle: All relevant general static and dynamic aspects, i.e. all rules, laws, etc., of the universe of discourse should be described in the conceptual schema. The information system cannot be held responsible for not meeting those described elsewhere, including in particular those in application programs. Conceptualization principle: A conceptual schema should only include conceptually relevant aspects, both static and dynamic, of the universe of discourse, thus excluding all aspects of (external or internal) data representation, physical data organization and access as well as all aspects of particular external user representation such as message formats, data structures, etc. Some justifications and remarks may amplify the two principles: Most of the general rules of the universe of discourse are currently described in application programs, if they are described at all. Sometimes such rules are referred to by the term "validation rules" in the current EDP jargon. In many situations, more than one update program operates on the same data base. In this case a single rule has to be described in each application program which may affect a concrete physical representation of that part of the information base covered by the rule. This results in a redundant representation of the rule, which is source for inconsistency among the various "copies" of the rule. It is obvious that due to such redundancy it is difficult to control, verify, and maintain a set of interrelated rules. It is even more difficult to detect contradictions in such a set of dispersed rules. User update and query languages are becoming available which provide the (end) user tools to express his requests directly to the machine as opposed to a professional application programmer who writes the program on behalf of the employing user. One does not need too much imagination to think of situations where a user "has no time" to program the rules in his direct instruction of the machine. This may result in a stream of messages entering the information base that should not enter. The result is information base pollution. If all rules are expressed in the conceptual schema, and therefore are controllable by system functfons independent of users or application programs, then there is no basis for such pollution. If all general rules of the universe of discourse are expressed in the conceptual schema, then it is (much) easier to extend or modffy these rules in a controlled way as compared to the situation where the rules are scattered over several application programs. The need for such extension and modification is a normal thing to expect in practice. It is easier to understand and to teach information systems design where all rules are in the conceptual schema, because of the absence of the previously mentioned kind of redundancy. A conclusion of the 100 Percent Principle is that a conceptual schema language must be able to describe any set of general rules, etc. of the universe of discourse in a conceptual schema. In order to achieve this, some of these rules may be described in a procedural way while others may be described in a declarative way. This depends on the present state of the art of formal languages. The Conceptualization Principle says that the conceptual schema must only and exclusively include conceptually relevant aspects of the universe of discourse. Aspects, constructs, or distinctions, which refer to other components or factors of the information system are not allowed to be part of the conceptual schema. Such aspects are, for example: - Representation aspects of data in the user views; - Aspects of machine efficiency and physical data organization; - Organizational aspects of the information system. By focussing on the conceptually relevant aspects, the conceptual schema design process is relieved of the burden of computer implementation aspects. The design process is in this way significantly simplified. More precise and accurate conceptual schemata will be the result of the conceptual schema design process. Moreover the design processes of the (external) application programs and the (internal) data storage routines will also be simplified as these can be limited to the external and internal representation aspects respectively. Furthermore, by avoiding aspects of machine efficiency and physical data organization in the conceptual schema, operations at the user interfaces can be made completely independent of those "interna1" aspects. Thus, reprogramming of applications in case of physical data reorganizations can be avoided, due to the "isolating" effect of the conceptual schema. Finally, the observation of this principle also results in better evolution, simply because there is no burden of conceptually irrelevant aspects which could increase the reprogramming effort in cases of conceptual schema changes.