For the author his case against RD is not for the implementation. At the end of the articles he calls for LR reference grammars (I imagine both for fuzzy testin

For the author his case against RD is not for the implementation. At the end of the articles he calls for LR reference grammars (I imagine both for fuzzy testing and documentation) to ensure that your perfectly usable RD parser with nice error messages does not miss some edge cases.