The ForkParser now builds the parser in the parent process and serializes it to the child process. It would be neat to make it easier for users of the ForkParser to depend solely on tika-core and put all of our dependency nastiness in a separate directory that will be used by the the fork server (child process) to build the underlying parser.
This would allow, e.g. Solr, to point to a directory with the tika-app.jar and remove all of our dependencies (except tika-core) from their dependencies.
I propose that we allow users to initialize ForkParser with a Path that contains all the jars necessary to build the Parser, and, optionally, a ParserFactory.