Titus

@ChristianMurphy it didn’t work before (https://redirect.github.com/mdx-js/mdx/blob/4235a925bd5bd0f9375225a6d0e06a9bafaedf5b/.github/workflows/website.yml#L10-L11). Was your review because you now think it would work, or without that knowledge?

> Didn't it just run and work though? It did **not** just publish to the website 😉 I tried again, and it failed. So, reverted > It looks like there...

These things are related; using `codePointAt`/`fromCodePoint` would break other places. Basically all of markdown doesn’t care, it only cares about ASCII. Except for this case. I think changing to code...

> Indeed markdown-rs seems to handle documents as the form of u8 sequences in other parts. Indeed! Being able to index into strings, whether the way JS does so, or...

So, why do you think the value you use for `Expected behavior` is correct? CM.js gives *actual*: https://spec.commonmark.org/dingus/. GH (so likely also CM in CI) does something else: a**a«**a a**»a**a...

Ah, that is sensical indeed; they are often behind. As you are using these characters, and there is so much difference across vendors, perhaps this makes sense as a test...

That’s not the job of an AST. No AST persists whitespace. You are trying to do things with parse5 which it can never do. Now, the root question of this...

First, there should probably (at the top?) be something about `→` representing a tab in this document. I would personally recommend a control picture instead: https://en.wikipedia.org/wiki/Control_Pictures. The character `␉` could...

Hmm, this doesn’t have anything to do with block quotes, from what I understand? What’s going on in these two cases is that the start condition of HTML kind 6...

Right, but that’s something else. The inline rules/algos have nothing to do with how (block quotes and) HTML interact.