tion and flow-control services provided by these products, message-translation services—which constitute another form of information integration—are also needed.

A typical message-translation scenario in e-commerce enables a small vendor (say, Pico) to offer its products through a large retail Web site ( Goliath). When a customer buys one of Pico’s products from Goliath, it sends an order message to Pico, which then has to translate that message into the format required by its order-processing system. A message-mapping tool can help Pico meet this challenge. Such a tool offers a graphical interface to define translation functions, which are then compiled into a program to perform the message translation. Similar mapping tools are used to help relate the schemas of the source databases to the target schema for ETL and EII and to generate the programs needed for data translation.

Object-to-Relational Mappers. Application programs today are typically written in an object-oriented language, but the data they access is usually stored in a relational database. While mapping applications to databases requires integration of the relational and application schemas, differences in schema constructs can make the mapping rather complicated. For example, there are many ways to map classes that are related by inheritance

into relational tables. To simplify the problem, an object-to-relational map-per offers a high-level language in which to define mappings. ²³ The resulting mappings are then compiled into programs that translate queries and updates over the object-oriented interface into queries and updates on the relational database.

Document Management. Much of the information in an enterprise is contained in documents, such as text files, spreadsheets, and slide shows that contain interrelated information relevant to critical business functions— product designs, marketing plans, pricing, and development schedules, for example. To promote collaboration and avoid duplicated work in a large organization, this information needs to be integrated and published. Integration may simply involve making the documents available on a single Web page (such as a portal) or in a content-management system, possibly augmented with per-document annotations (on author and status, for example). Or integration may mean combining information from these documents into a new document, such as a financial analysis.

Whether or not the documents are collected in one store, they can be indexed to enable keyword search across the enterprise. In some applications, it is useful to extract structured information from documents, such as cus-

tomer name and address from email messages received by the customer-support team. The ability to extract structured information of this kind may also allow businesses to integrate unstructured documents with preexisting structured data. In the example above, the auto manufacturer wanted to link transactional information about purchases with emails about these purchases in order to enable better analysis of problem reports.

Portal Management. One way to integrate related information is simply to present it all, side-by-side, on the same screen. A portal is an entire Web site built with this type of integration in mind. For example, the home page of a financial services Web site typically presents market prices, business news, and analyses of recent trends. The person viewing it does the actual integration of the information.

Portal design requires a mixture of content management (to deal with documents and databases) and user-interaction technology (to present the information in useful and attractive ways). Sometimes these technologies are packaged together into a product for portal design. ¹¹ But often they are selected piecemeal, based on the required functionality of the portal and the taste and experience of the developers who assemble it.

 

figure 2. screenshot of a mapping tool.

core technologies

Extensible Markup Language (XML). In any of the scenarios noted here, an integrated view of data from multiple sources must be created. Often any one of the sources will be incomplete with respect to that view, with each source missing some information that the others provide. In our example, the emails are unlikely to provide detailed information about the dealerships, while the relational data might not have the problem reports. In XML, a semi-structured format, each data element is tagged so that only elements whose values are known need to be included. This ability to handle variations in information content is driving EII systems to experiment with XML. ²²

This flexibility makes XML an interesting format for integrating information across systems with differing representations of data. In some integration scenarios, it may not be