Perspective for Database Preservation

Introduction

The approaches of database preservation described in (Buneman & Christophides, 2007) do not consider the need to preserve database concepts. In database models a label denotes a class/a set of objects/ and connotes a corresponding concept; as an example, the label “department” denotes the class “department” and connotes the concept of department. Concepts and classes of objects are modeled in a similar way. In concept theory (Kauppi, 1997; Palomäki, 1994), concepts are introduced at the intensional/concept level, which is the level of human thinking, whereas classes/ set of objects are introduced at the extensional/set-theoretical level, which is the level of computer science. A distinction is made between concepts and classes of objects.

At the intensional level, concepts are modelled in terms of information contents of concepts, and concept structures are defined through an is-in relation, called intensional containment relation, (Palomäki & Kangassalo, 2012). Correspondently, at the extensional level, an is-a relationship is exploited to model database specialization hierarchies. Oriented acyclic graphs of classes in is-a relationships supported by Semantic Data Models (Cardenas & McLeod, 1990) or Enhanced Entity-Relationships models (Elmasri & Navathe, 2000), called Conceptual graphs, are similar to oriented acyclic graphs of classes in is-a_o relationships supported by object database systems, called Logical object graphs (Locuratolo, 2013). The following properties hold:

• •

  Is-a relationship: Each object instance can belong to any class of a conceptual graph. This property enhances flexibility in modeling changes occurring in the real life.

• •

  Is-a_o relationship: Each object instance belongs to one and only one class of the logical object graph. This property enhances efficiency in accessing and storing objects.

The difference between an is-in relation and an is-a/ is-a_o relationship can be explained as follows: in a conceptual model using an is-a relationship as its taxonomical link, objects to be modeled are presupposed to exist; vice-versa, in a concept structure using an is-in relation, the existence of objects falling under a concept is not presupposed. In a conceptual model using an is-a relationship, objects can be added to classes or removed from them while the concepts of the corresponding classes remain unchanged.

Original research based on concept theory has recently been proposed for database preservation (Locuratolo & Palomäki, 2013). By ontology for database preservation we will refer a structure where all and only the concepts related with a conceptual graph are retained; the logical coherence among these concepts is maintained and a mapping from the concept structure to a corresponding logical object graph is provided (Locuratolo & Palomäki, 2013). This concept structure is characterized by the following properties:

• •

  Results in leaves which are incompatible concepts.

• •

  Encloses all and only the concepts related to an initial concept structure.

• •

  Encloses all and only the intensional inclusion relations among concepts.

• •

  Is mapped to a logical object graph.

In ontology for database preservation, concepts are mapped to classes of objects, but separated from them, thus, in addition to the database preservation results achieved in (Buneman & Christophides, 2007), the further result of preserving all and only the concepts included in oriented acyclic graphs of object classes is considered.