XML can be thought of as the "universal serialization of data." It provides a flexible, open approach for modeling data and sharing messages among business partners (or systems) in a consistent manner. XML provides the ideal solution to messaging in a B2B e-commerce infrastructure since it enables a loosely coupled design that can significantly lower a partner's barrier to entry.

While most users of XML utilize its hierarchical nature to define data, this article discusses possible approaches for eliminating redundant data within XML messages by employing the features of ID and IDREF to define a more relational approach to defining the data.

Reporting Data Is Redundant
Developing a common vocabulary for a single enterprise or a large B2B marketplace can be a daunting task. With XML the challenge of modeling commonly utilized data models in an open manner can be resolved. It's based on a formal W3C Recommendation and has extensive support from both users and IT vendors, thereby providing a stable, future-proof foundation for business applications. The technical and business benefits of utilizing an open, standards-based XML strategy are immense. Whenever possible, an enterprise's existing data models and semantics should be utilized to construct the markup language.

There are, however, several issues that must be addressed prior to effectively representing an organization's data models in XML. Common reporting and aggregate data sets frequently contain repetitive data such as name and address. These redundancies are usually carried over into the resulting markup. This approach leads to repetitive data structures - the result of which is poorly designed markup that doesn't make use of reusable data models. XML's ID and IDREF tokenized attributes enable designers to leverage relational modeling concepts and avoid creating redundant data structures.

ID is a specially defined attribute that uniquely represents an instantiation of an element within an XML message. The value of an ID attribute is similar to a primary key value in a relational database table in that no two ID attributes (for a given element) can have the same value. IDREF is another special XML attribute that references a previously defined ID value. IDREFS is similar to IDREF, but can point to several previously defined ID values, each separated by a space. Listing 1 displays a simple example of how ID, IDREF and IDREFS are typically used in an XML document.

As illustrated in Listing 1, ID, IDREF and IDREFS can be used to establish relationships within the data. They can also be used to represent data structures without utilizing nested elements. This approach enables developers to model data that lacks definite structure (such as a result set from a user-defined query).

Based on these concepts, a set of redundant elements can be effectively modeled using ID and IDREF attributes. A sample document with redundant elements appears in Listing 2. As this listing clearly illustrates, the Key, Firstname and Lastname elements appear in several locations throughout the document.

Listing 3 illustrates how this data can be modeled in a more effective manner using ID, IDREF and IDREFS.

Note that the code in Listing 3 establishes a data structure (Keys) that can be reused elsewhere in the document. This design reflects the relational approach to defining and utilizing translation tables to avoid many-to-many relationships (as seen in Figure 1).

Processing XML Using Xpath and ID/IDREF
Today's XML parsers don't make use of the ID/IDREF functionality as defined in this article. However, the W3C XPath specification ( www.w3.org/TR/1999/REC-xpath-19991116) provides facilities for identifying and selecting nodes based on their IDs. Additionally, IDREFs can automatically be retrieved from identified nodes and supplied directly as input into the ID function.

The XPath specification defines a set of core functions that must be available to all implementations of XPath processors. One of these functions operates over an XML document to provide a node-set result based on the id function. The id function takes a single string parameter identifying the node that defines the unique ID. Using the example in Listing 1, we could use the following XPath statement to retrieve the supervisor for a particular employee:

id("456") returns the SUPERVISOR node for Debbie Hamel

Conclusion
As you can see by the differences in Listings 2 and 3, using XML to define data hierarchically can produce highly redundant elements. By using the method described in this article, it's possible to mix hierarchical and relational techniques within the same document, which results in a more usable and reusable XML document. In addition, this document will contain clearly recognizable elements that can be employed to describe information relative to data both within and external to the XML document.