Richard Gibson

> I'd rather have all keys without quotes, regardless of whether they are identifiers or not. That would violate "[The output of gron is valid JavaScript](https://github.com/tomnomnom/gron#usage)", see https://tc39.es/ecma262/multipage/ecmascript-language-expressions.html#prod-MemberExpression and https://tc39.es/ecma262/multipage/ecmascript-language-lexical-grammar.html#sec-identifier-names-static-semantics-early-errors...

What if a property name includes the delimiter sequence?

I just noticed that this was still open. Any plans to land it?

Another request for this functionality: https://github.com/unicode-org/message-format-wg/issues/501#issuecomment-1794517256

Well, it is conceivable that the future evolution of the language will lead to a place where heavy use of receivers is commonplace. The existing primary indirect invocation facilities—`Function.prototype.apply`, `Function.prototype.call`,...

> Function.prototype.apply/call and Reflect.apply are special helpers, which are very rarely sensible precedent to use for other methods – https://matrixlogs.bakkot.com/TC39_Delegates/2024-04-18#L24 For the record, color me convinced.

This would be fixed by merging #48.

> But, an adequate mitigation for this specific case might be to augment `@endo/evasive-transforms` to change all occurrences of `x-- > y` to `(x--, x > y)`. That's not equivalent...

It looks to me like initialization of the marshal package and use for marshalling/unmarshalling CapData currently has three firm dependencies that are satisfied by `@endo/init`: * a global `harden` function...

@ljharb Done (both).