Currently, the doclet code uses DocletConstants.NL to represent an OS-specific newline -- i.e. System.getProperty("line.separator").

It would be better if the code used \n internally, and converted to using NL on output, perhaps by scanning content and using println on each line.

This would make the code less cumbersome, and would make the code catch all newlines (even those in user content) and translate them appropriately.