Configure: teach the tokenizer to handle other separators than spaces