Editorial: remove %Generator% and %AsyncGenerator% in favor of dotted forms

[ljharb]

Fixes #2023.

#2056 will need to be rebased after this lands.

[ExE-Boss]

I think we should keep the <dfn>s in the dotted forms.

---

Also, this will need to be rebased on top of #2059.

[jmdyck]

I agree that it's redundant to say it's both:

• the initial value of the "prototype" property of %GeneratorFunction%, and
• the intrinsic object %GeneratorFunction.prototype%.

But I'm wondering why you chose to keep the first and remove the second, rather than vice versa. (We've established the %Foo.bar% notation, so why not use it here?)

Ditto the deletions under:

• "The AsyncGeneratorFunction prototype object"
• "The Generator Prototype Object"
• "The AsyncGenerator Prototype Object"

[ljharb]

%GeneratorFunction.prototype% only works because the prototype property is already defined; in other words, the first of those two is the one that's actually required. The second one isn't necessary, I suppose, but is nice for clarity.

[jmdyck]

If the spec says "X is %Foo.bar%", then by definition, that means that if I were to access the 'bar' property of '%Foo% prior to any ECMAScript code being evaluated, I would get X. To me, this seems equivalent to saying that X is the initial value of the 'bar' property of %Foo%. You're saying these aren't equivalent?

[ljharb]

@jmdyck i'm saying that relying on that kind of inference isn't something I'd prefer; %Foo.bar% works because it's the initial value of bar on Foo, which imo should have to be defined somewhere.

[jmdyck]

Isn't 25.2.2.2 Generator.prototype enough of a definition?

[ljharb]

25.2.2.2 there doesn't actually have an explicit link that I can see to 25.2.3, "the GeneratorFunction prototype object"; 25.2.3 has "is the value of the "prototype" property of %GeneratorFunction%." which is the link.

[jmdyck]

I'm confused, so I'll back up. Earlier, I asserted the equivalence of the two statements:

• X is %Foo.bar%
• X is the initial value of the bar property of %Foo%.

You didn't deny that they were equivalent, but you said that the first relies on a kind of inference (on the part of the reader, presumably) that you didn't prefer. I don't really understand that, but I can probably accept it as an editorial preference.

So, to clarify the specifics of this preference, 2 questions:

1. Would it make a difference for you if the statement were the other way round:
   %Foo.bar% is X ?

2. Do you expect to eventually remove all statements of the form The initial value of `Foo.prototype` is %Foo.prototype%., or do any of them serve a purpose?

[ljharb]

I'm saying that they're not equivalent unless the second one exists.

[jmdyck]

To clarify, when you say "the second one exists", do you mean "somewhere in the spec establishes that %Foo% has a bar property"?

[ljharb]

Something that establishes that Foo has a bar property, and that the value of that property is "the object in question" (ie, "the Array prototype object", or whatever)

[ExE-Boss]

Based on your suggestion in #2059 (comment), the wording:

<li>is the initial value of the *"bar"* property of %Foo%.</li>

Is redundant and should be replaced with the simpler:

<li>is <dfn>%Foo.bar%</dfn>.</li>

Which has the added bonus of supporting auto‑linking for the dotted intrinsics.

[ljharb]

Yes; but this PR shouldn't be updated to that wording until #2059 lands.

[ExE-Boss]

This wasn’t rebased correctly.

See also:

• Editorial: Add auto‑linking for dotted intrinsics #2041

---

P.S.: Please keep the emu‑xref in the definition of the Generator Function Prototype Object:

(See <emu-xref href="#figure-2"/>)

[ljharb]

updated

[littledan]

This change seems to increase verbosity and make it harder to tell what the spec is talking about. What is the motivation?

[ljharb]

@littledan overall, the editors decided that we should only have one preferred way to refer to every intrinsic, and that the shortest (and dotted, when applicable) form would be preferred. Regarding %Generator% and %AsyncGenerator%, see #2023 (comment); during the editor call the consensus was to remove these forms, particularly because we found the names misleading.

[jmdyck]

This PR leaves dangling references...

• to %Generator% in Properties of the GeneratorFunction Prototype Object, and
• to %AsyncGenerator% in Properties of the AsyncGeneratorFunction Prototype Object.

[ljharb]

All updated, including →

[ExE-Boss]

This doesn’t currently preserve auto‑linking for %GeneratorFunction​.prototype .prototype% and %AsyncGeneratorFunction .prototype​.prototype%.

See also:

• Editorial: Add auto‑linking for dotted intrinsics #2041

[ljharb]

I'm not convinced we want to preserve auto-linking for something that's not actually a well-known intrinsic anymore.

[ljharb]

Now that #1948 has consensus and has been merged, this PR is no longer normative.

[bakkot]

@littledan I am personally in favor of this change because I cannot for the life of me keep track of which is %Generator%, and which is %GeneratorFunction%, whereas %GeneratorFunction% vs %GeneratorFunction.prototype% is unambiguous.

[bakkot]

This and many of the subsequent changes seem unrelated to this PR; is this a rebase conflict? If it's intentional, I'd ask that these changes go into their own PR.

[ljharb]

It was intentional to make everything consistent. I can pull them out into another PR tho.

[ExE-Boss]

I believe that it should probably be done in its own PR/commit.

[ljharb]

@bakkot i extracted it to #2131 and rebased this one on top of it, so you can review just the commit for this PR separately. If you'd prefer to wait to review this one until after #2131 lands, that's fine too.

[jmdyck]

Changes look okay to me.

[ExE-Boss]

For consistency with current definitions, the <dfn> for the intrinsic name should come before “is an ordinary object”.

[ExE-Boss]

This shouldn’t yet close #2023, that should only be done by #2056.

[ljharb]

@ExE-Boss the editorial convention is established by this PR; removing the legacy ones entirely is a cleanup step after that's resolved.