The DITA-OT processing pipeline is implemented using Ant. Individual modules within the Ant script are implemented in either Java or XSLT, depending on such factors as performance or requirements for customization. Virtually all Ant and XSLT modules can be extended by adding a plug-in to the toolkit; new Ant targets may be inserted before or after common processing, and new rules may be imported into common XSLT modules to override default processing.
The XSLT modules use shell files. Typically, each shell file begins by importing common rules that apply to all topics. This set of common processing rules may in turn import additional common modules, such as those used for reporting errors or determining the document locale. After the common rules are imported, additional imports can be included in order to support processing for DITA specializations.
For example, XHTML processing is controlled by the xsl/dita2xhtml.xsl file. The shell begins by importing common rules that are applicable to all general topics: xslhtml/dita2htmlImpl.xsl. After that, additional XSLT overrides are imported for specializations that require modified processing. For example, an override for reference topics is imported in order to add default headers to property tables. Additional modules are imported for tasks, for the highlighting domain, and for several other standard specializations. After the standard XSLT overrides occur, plug-ins may add in additional processing rules for local styles or for additional specializations.
Java modules are typically used when XSLT is a poor fit, such as for processes that make use of standard Java libraries (like those used for index sorting). Java modules are also used in many cases where a step involves copying files, such as the initial process where source files are parsed and copied to a temporary processing directory.