The generate-debug-attributes parameter determines whether debugging attributes are generated in the temporary files. By changing the value to false, DITA-OT will no longer generate the xtrf and xtrc debug attributes. This will make it more difficult to track down the source file location from which a given issue may have originated, but it will reduce the size of the temporary files. As a result, XML parsing will take less time and overall processing time will be reduced.

DITA-OT keeps topic and map files as separate files and processes each file multiple times during preprocessing. Thus reading from disk, parsing XML, serializing XML, and writing to disk makes processing quite I/O intensive. Use either an SSD or a RAM disk for temporary files, and never use a temporary directory that is not located on the same machine as where the processing takes place.

For all but extremely large source sets, the JVM will not have enough time to warm-up. By reusing the same JVM instance, the first few DITA-OT conversions will be “normal”, but when the JIT starts to kick in, the performance increase may be 2-10 fold. This is especially noticeable with smaller source sets, as much of the DITA-OT processing is I/O intensive.

DITA-OT 2.0 to 2.3 require Java 7, and DITA-OT 2.4 and newer require Java 8. However, using a newer version of Java can further reduce processing time.