I wrote the Erlang Markdown interpreter https://github.com/hypernumbers/erlmarkdown Real-life Markdown almost always requires that something be rendered twice: * on the server * in the client preview The sane way to make this easy is for the javascript version to be the dominant one - and all the servers-side language version to test themselves the same way: if I start from the same markdown do I produce the same whitespace compatible output as the javascript parser. When I wrote erlmarkdown I first implemented my take on the spec - then realised I needed to switch to tracking a reference javascript version - which didn't exist/wasn't maintained/etc, etc Then I discovered that the whitespace output of the javascript parser was all over the shop, random linebreaks and stuff - and it turned out to be super-hard to backport that into my hand-written, look-ahead parser. Most of the whitespace difference was just html-farts of no consequences - but the inability to produce whitespace equivalent meant I couldn't write simple asserts like: * =?equals(Got, Expected) It was generally a real pain. PS I also think that markdown implementations should accept all html and xml tags from an arbritrary implementation-time whitelist. PPS They need to accept unicode input as well.