A TARGET : THE ACS

P.M. STOCKER* and R. CANTIE)**

*UNIVERSITY OF EAST ANGLIA, NORWICHNR4 7TJ, UNITED KINGDOM **MATRA, TOULOUSE, FRANCE Abstract 2. The Abstracted This extended abstract describes a The ACS is based upon two logical structures, (ACS) intended to serve as a target for the classical set and an indirect functional conceptual schema languages and as a source for mechanism called the element set. implementations. It indicates some of the uses to which it has been put. All data values are defined in terms of data- item types; a dataitem type is a unitary item of 1. Introduction data which is not further divisible for any pur- Over the past decade two related trends have pose; it is a logical item, e.g. 'salary in become apparent. The first, in the dollars'; information concerning representation area, is to capture some of the semantic meaning (internal and external) is additional to the ACS. of the data, and to store this also in the data- Information concerning comparability, that base by means of a 'conceptual schema'. The 'salary in dollars' may logically be compared with second, in the area of programming languages, is 'tax in dollars', is part of the ACS. the move to non-procedural programming, complex The element set provides the functional data types whose structure is not revealed to the association between dataitems. An element set user, constraints embedded in data types and so type is a set of function/dataitem type combina- on. Both demonstrate the transition to the tions. For brevity the function/dataitem type situation where internal structure and combination will be called a property and, when a representation within the computer are no longer value of the dataitem is associated with it, we the concern of the programmer, who works shall refer to a property value. A graphic repre- increasingly at a logical level. sentation is shown in Figure 1, fi, di and pi A conceptual schema language has two roles, refer to functions, dataitems and properties one is that of providing a language which allows respectively. its user to make the transition from the model of the universe of discourse to defined, p1 fl -dl representable data items. This role serves to f-d define the relationships between the data items element p2 2 2 set f-d of concern and the constraints upon them, and p3 1 3 also provides a supportive structure which helps f-d the user to resolve his concept of the transition. p4 E 3 1 The second role is to capture and store these Figure 1 relationships and constraints in a form which relates to the purpose of providing a basis for Elements of the same element set type are an implementation of the database. This abstract contained in an element set. Thus an element is concerned with the second role, that is, with occurrence is a tuple of property values (e.g. the mechanism for retaining the information dl, d2, d3, dl in Figure 1). Properties are captured by the conceptual schema language rather divided into identifying properties (identifiers) than the language itself. . and non-identifying properties. An identifier uniquely determines an element within an element In the transition, through the various layers, set, i.e. only one tuple within the set takes from universe of discourse to stored bytes of this value. basic data, explicit information disappears. It is either stored within a layer, to be reincor- The element set mechanism, therefore, has the porated on the outward path, or it becomes following result. There is a functional rela- implicit in a data structurel. This abstract tionship between the values of an identifying proposes a small set of structural forms which property and the values of each other property of This mechanism is suffice to contain the information obtained the element set type. through the conceptual schema language, the set equivalent to the Nijssen 'bridge', expressed in Identifying is called the Abstracted Conceptual Schema (ACS). n-ary rather than binary terms. It is intended to serve as a target for properties relate to LOTS, non-identifying conceptual schema languages. properties to NOLOTS.

309 The element set is one way of capturing the entities and relationships, since both map into relationships established by the conceptual similar element structures. Moreover, the schema language. Relationships are also implied 'special role' of time is transmuted into where two element sets have a common property, conventional functional relationships. (weaker) a common dataitem type. or, 5.3 It forms a basis for a comparison between 3. Additional Structures schemata obtained from conceptual schema 3.1 A combination of properties may be used, languages and schemata describing implementation, called a compound property, and a corresponding e.g. an ADAPLEX schema with the CODASYL DDL for identifier defined. its implementation. In this context it has been adopted as the basis for a heterogeneous 3.2 The transmission of properties through a distributed database project, PROTEUS. hierarchy (or network) of element sets is incorporated (using matched identifying 5.4 It has proved convenient as a basis for a properties). This provides the structure number of varied current research projects, e.g. necessary to capture 'roles'. unification of schemata which overlap, 'intelligent' aids to form construction. 3.3 The constraint mechanism is implemented by means of classic set theory. The set of values 5.5 It may serve as a point of discussion on taken by any property (or combination of possible open network architecture protocols, properties) may be formed into a set of data- since it allows information interchange in a item values (or tuple values), called a value relatively uncoloured and non-committed form4. set. A value set may be defined on one element 5.6 The relational form of representation is set and then used to constrain the allowable highly convenient as one of the inputs required domain of property values taken in another for automatic implementation. element set, by means of set identity or containment operators. The normal operations of 6. Information Representation set theory may be used to construct new value It would, in many instances, be advantageous sets, from values of comparable dataitem type. if a model of given information content trans- 4. Representation of the ACS formed into the same ACS structure, irrespective of the conceptual schema language used in the The ACS is a number of data structures, not process. The ACS itself has semantic rules. Ad- a language. It could be represented in many ditional rules may be added to control the split- ways. The representation which has been used so ting of element sets by definition as against the far is in terms of relations. It is provision of a property which achieves the same representable (for site to site transmission, for effect. So it is possible that the ACS might be instance) in terms of 15 relations; of these 9 constrained into unitary, orthogonal constructs. are trivial in the sense that they correspond to These comments apply to structure; the manner in the name declarations of a programming language, which the conceptual schema languages make use of dataitem type names, function names, etc. The names prohibits equivalent naming. relations which carry structural content are: i) definition of properties of element sets References ii) definition of identifying properties of element sets 1. P.M. Stocker. "Canonical Schemata, Canonical iii) definition of value sets Queries and Standardisation", Infotech State- iv) application of value sets as constraints of the Art Report, Series 9, Number 8, v) construction of value sets by operators Database. vi) specification of 'role' networks. Information Systems Design Methodologies: 5. Utility of the ACS A Comparative Review. Proc. IFIP TC8 Working The presentation of the conceptual schema as Conference on Comparative Review of Infonna- data rather than procedure appears to have at tion Systems Design Methodologies, least tactical merit. Noordwijerhout, The Netherlands, May 1982. 5.1 It has been assessed as a target for a number M.J. Smith, P. Bernstein, U. Dayal, of conceptual schema languages discussed at the N. Goodman, T. Landers, K.W.T. Lin and "Multibase-integrating Heterogeneous CRIS-1 conference', NIAM, CIAM and ACM/PCM; an E. Wong. automatic both-way mapping with ADAPLEX3 is Distributed Database Systems", AFIPS, operational, as is a mapping from CODASYL DDL. Chicago, 1981. No difficulties were met in these mappings and Concepts and Terminology for the Conceptual the set based constraint mechanism was found to Schema and the Information Base. be capable of capturing more procedurally J.J. van Griethuysen. IS0 TC97/SC5/WG3 - expressed constraints than had been anticipated. N695, 1982. 5.2 Ihe structural form avoids arguments which arise in the supportive role of conceptual schema languages concerned with distinctions between

310