I've re-written my DiffMk application in Java. This version should be more accessible (if installing a Java environment is easier for you than installing Perl) and faster. It is definitely “better” in terms of features.
Note, however, that it employs a different implementation of the mapping from XML to arrays and back for purposes of using the underlying diff code, so it does not, and cannot, produce exactly the same results as the Perl version.
This is a really beta. You have been warned. You will probably have trouble, but if you shout for help, I'll do my best.
I've moved this project to SourceForge.