5.1 Introduction

Today's data-intensive applications bring a challenging collection of requirements to data management. These requirements include

• Accessing data from heterogeneous data sources. Data are stored not only in traditional database management systems (DBMSs), but also in simple files, index-sequential files, desktop databases, spreadsheets, project management tools, electronic mail, directory services, and spatial and multimedia data stores. Recently, the explosive growth of the World Wide Web (WWW) has transformed this medium into a vast repository of information and knowledge. All these data are usually heterogeneous in their native forms.
• Specialized data management. Given that heterogeneous data sources are the norm, each data source tends to have different strengths and weaknesses in the areas of data model, query language, transaction, and indexing support. Applications want to access data directly from their source of origin and exploit their data stores' native capabilities. For example, messaging data stores have a tree-structured name space and manage collections of heterogeneous objects such as messages, appointment, and contacts. Spatial and image data stores require support for geometric data types, indexes, and query language.
• Different data navigation models. The environments where the application runs may have different requirements for the way data are navigated. For example, an application running in a middle tier may operate on a small subset of data cached entirely within the application space and provide bidirectional or positional access to data. An application running in the database usually works on large data sets and tends...