Host: Bob Schloss

Web Services depend on XML documents that can be understood by the parties (or applications) that wish to do business. But where do these documents come from?

"Document Engineering" is emerging as a new discipline for specifying, designing, and implementing these XML documents and the business processes that produce and consume them. The methodology of Document Engineering is a "soft" one, a "document-centric" perspective on the classical Analyze/Design/Refine methodology.

Document Engineering is a soft methodology because distributed computing and web service architectures make it unreasonable for any enterprise to impose a modeling methodology on another one. The principle of "loose coupling," fundamental in this new architectural vision, is that for one business to interact with another one, it needs only to understand the interfaces offered by the other business and the information that is exchanged to satisfy those interfaces. This understanding is exposed in modeling artifacts. There is no need to understand the process by which the other business modeled its services or system.

Doing business requires both "publication-like" document types like brochures and technical manuals and "transactional" documents like purchase orders and invoices, so Document Engineering needs analysis and design methods that work for both ends of this "Document Type Spectrum." The traditional approaches of document analysis and data modeling come from different disciplines and use different tools, terminology, and techniques. Document Engineering can emphasize what these analysis and design approaches have in common rather than highlighting their differences:

• Identifying the presentational, content, and structural components and defining their relationships to each other
• Identifying "good" content components
• Designing, describing, and organizing components to facilitate their reuse
• Assembling hierarchical document models that organize components according to the requirements of a specific context for information exchange.

Good Document Engineering practice emphasizes the reuse of existing models or patterns, many of which are encoded at the implementation level in the form of EDI and XML vocabularies. Other patterns are at more conceptual levels in terms of common business processes or in patterns for the organization of activities between businesses using supply chains, marketplaces or hubs.

The "Reuse Matrix" organizes four levels of abstraction or model types, varying from highly abstracted conceptual models to specific instances of documents. Against the levels of abstraction we also have the depth at which we describe the patterns of document exchanges that are required by each model. These descriptions range from patterns of document exchange viewed at the “business to business” level, such as “vendor managed inventory” or “build to order,” to patterns of exchange from the perspective of a single business (such as a RosettaNet Partner Interface Process), to the most granular perspective that shows the re-use of components such as the common Party structure in XML schema libraries like UBL.

The common goals but different approaches of document analysis and data modeling can be depicted in the Reuse Matrix, as can the relationship of these two approaches to business process analysis. Document analysis typically starts with instances of data components in document artifacts, while data modeling generally starts with logical models of objects and associations. Business Process analysis typically starts abstract views of high level business models. Regardless of their starting point, all three of these approaches needs to reach the same place in the Reuse Matrix, the point at which most reusable patterns exist, where patterns have enough abstractness to be reusable, but enough concreteness to be prescriptive.

In practical terms this covergence depicts one of the key principles of Document Engineering - that models for documents and business processes need to be developed at the same time, with the same care, and to compatible levels of detail.

Document Engineering is being taught to graduate students at UC Berkeley and has been successfully applied to problems of significant scale, including efforts to develop an enterprise data architecture for the e-Berkeley initiative and the Universal Business Language.

Speaker Bio:

Bob Glushko is an Adjunct Professor at the University of California at Berkeley in the School of Information Management and Systems. He founded or co-founded three companies, the last of which was Veo Systems in 1997, which pioneered the use of XML for electronic commerce before its 1999 acquisition by Commerce One. From 1999-2002 he headed Commerce One's XML architecture and standards efforts and became the "XML Evangelist" after being named an Engineering Fellow in 2001. He has a BA from Stanford, a MS (software engineering) from the Wang Institute, and a Ph.D. (cognitive psychology) from UC San Diego.