XML as a Pragmatic Choice for Domain-Specific Languages

The article presents a compelling case for reconsidering XML's role in modern software development, particularly as a domain-specific language (DSL) for complex, cross-platform specifications. The author, working on the IRS Tax Withholding Estimator (TWE), demonstrates how XML provides a robust foundation for representing the intricate logic of the US Tax Code.

The core argument centers on XML's declarative nature, which contrasts sharply with imperative programming approaches. When representing tax calculations, the XML-based Fact Dictionary clearly defines relationships between different tax values without specifying execution order. This approach offers significant advantages for auditability and introspection, allowing the system to explain how it arrived at any particular calculation—a crucial feature for tax software where transparency is paramount.

The author provides concrete examples comparing XML representations with equivalent JavaScript code. While the JavaScript version appears more concise initially, it quickly becomes unwieldy as dependencies grow, mixing implementation details (like array manipulation) with the core tax logic. In contrast, the XML representation maintains a consistent, declarative style throughout, making the relationships between values clearer and more maintainable.

A particularly insightful comparison examines how different data formats handle complex nested expressions. The author demonstrates that JSON struggles with arbitrary nesting, requiring verbose type declarations for each element. XML, with its natural hierarchical structure and attributes, provides a more elegant solution that can be more easily read and edited by humans.

The article also addresses the practical benefits of XML's mature ecosystem. The author leverages XPath and standard Unix tools to build powerful debugging and exploration utilities with minimal effort. This "universal data representation" quality allows developers to work with the data in whatever language or tool best suits their current task, without being locked into a particular implementation.

While acknowledging that other formats like s-expressions or Prolog might offer more elegant syntax, the author argues that XML's tooling support and universality make it the "cheapest" DSL option in terms of development cost and efficiency. This pragmatic perspective is refreshing in a field often preoccupied with finding the "perfect" solution rather than the most effective one.

The IRS's open-source approach to the Fact Graph represents an important contribution to public sector software, demonstrating how declarative XML specifications can make complex systems more transparent, maintainable, and accessible to public scrutiny. This case study serves as a valuable reminder that the best tool for a job isn't always the newest or most fashionable one, but rather the one that best addresses the specific requirements of the problem domain.