Clarify if `'` is allowed (specification-implementation mismatch)

[CodesInChaos]

The specification says "List of 252 character entity references defined in HTML4 standard" are allowed and links to at here for the list.

The html4 list does not include ', while xhtml/xml and html5 allow it. So I read the specification as disallowing '.

However the implementation used by the jsx demo does accept ' and renders it as '.

[magic-akari]

Hi! While investigating swc-project/swc#12123, we found this same ambiguity in SWC's JSX handling.

For this input:

const a = <p>&apos;&amp;</p>;

This led us to look more closely at the intended JSX semantics. As noted in this issue, the written JSX specification and the JSX demo appear to differ on whether ' is supported.

We see two possible ways to resolve this ambiguity:

1. Treat apos as a supported named reference. This matches the current behavior of Babel and TypeScript.
2. Treat apos as unsupported and preserve it as the literal text '. This follows the literal HTML4/252-entity reading of the JSX specification.

We found that #136 intentionally chose HTML4 named entities for portability, but it does not appear to discuss apos as an explicit exception. Meanwhile, the JSX demo accepts ', as noted in this issue.

Could you clarify the intended normative behavior?

• Is ' intentionally excluded, because the supported named-reference set is exactly the HTML4 list of 252 names?
• Or is apos intended as a legacy exception, in which case the specification should describe the 253-name set explicitly?

We would like SWC to follow the intended JSX behavior rather than choose based only on compatibility with another implementation.

CC @Huxpro, as the author of #136.

[CodesInChaos]

I prefer adding ' to the specification. It's very useful for escaping a ' inside a single quoted string.

And XML only has 5 named entities, 4 of which are already allowed in JSX, with the fifth being ' so XML compatibility can be added at minimal cost. If I had designed the spec, I'd probably only supported those 5 entities, since these are the characters having special meaning, and get escaped by minimal html encoders.