The current version is not particularly robust and fails when encountering bad or unexpected data. This should be resolved