Heterogeneous network distribution is as inevitable a fact of the enterprise environment today as XML is in the management of enterprise information.

In order to utilize the full potential of XML and provide for maximum flexibility, we must be able to access XML information in a fashion that is independent of how system components are distributed across networks, operating systems used, and programming languages. This is exactly what CORBA provides, but not without a price. Remote procedure calls are inherently expensive, network bandwidth is not free, and the difficulty of implementing a standard that has mappings to every major programming language can be complex. The use of XML data in such a distributed system adds the additional complication of providing efficient access to highly granular XML structures, such as DOM or grove nodes as well as providing appropriate feedback from server to client about XML processing, such as validation.

This paper presents our approach to using CORBA to expose XML information in a manner that provides a consistent, language-independent interface for programmers while providing full flexibility in determining the granularity and consistency of the information as accessed. The system we have developed allows for access across a spectrum of granularity vs. transmission efficiency, from direct access to individual nodes to bulk transmission of entire trees. The system also provides robust mechanisms for exposing XML processing information (e.g., error and status messages) to clients.

The paper begins with the problems inherent in distributed processing, especially with CORBA, the different ways that using XML in a distributed system exhibits those problems, how we have addressed those problems using the aforementioned interface implemented using the Python CORBA bindings, and the problems that have yet to be completely solved (primarily the flexibility and robustness provided at the cost of some API complexity).