| For relational database systems, normalization plays a major role in producing a good database design. In general, good design refers to schemas that avoid storing anomalies. It has been proved that some normal forms like BCNF and 3NF are well designed ones in the information theoretic sense.;Today, the number of applications using XML documents is increasing rapidly [41]. Like relational databases, XML research has also focused on better design. Functional dependencies and multivalued dependencies in relational databases have been extended to XML tree design. The main aim of this paper is to study design principles for recomposed XML trees and to prove its nested structure is a good representation.;I first develop a transformation algorithm for the XML tree structure to transform it into its corresponding flat and nested forms. During the transformation process, a tree may have operations such as nesting and unnesting applied. If the final structure is not affected by these operations, and if it is equivalent to the original one, then the structure is said to be unique up to the order of a tree. Any tree structure can be transformed into a unique order flat form.;I use an approach based on information theoretic concepts to measure the information content of the data. The measure has been applied to the relational model to prove the goodness of design. Previously information theoretic measures were applied to the flat model for XML designs. The motivation of my work is to apply this method in the nested model and show its quality.;Next, the properties of a recomposed tree are discussed. An XML tree can be represented by different structures. If any tree has an attribute with set values, the same attribute in other elements is not forced to have set values. They may have either singleton or set values. I use a theorem to compare the nested tree design with its equivalent flat design. Finally, a summary is given to show some important properties of a tree. |
