Editorial: remove legacy intrinsics

[ljharb]

All uses of these have been removed from HTML and 402:

• add links for dotted intrinsic form whatwg/webidl#898
• Use dotted intrinsic form from ecma262 whatwg/html#5655
• Use dotted intrinsic form from ecma262 whatwg/infra#316
• Editorial: use dotted intrinsic form ecma402#474

Closes #2023. Closes #2124.

[jmdyck]

The Well-Known Intrinsics clause has (what might be read as) two definitions of "well-known intrinsics":

• built-in objects that are explicitly referenced by the algorithms of this specification, and
• the objects listed in the table "Well-Known Intrinsic Objects".

Currently, these definitions are more or less in agreement, but with this PR, they very much disagree.

E.g., is %ArrayBuffer.prototype% a well-known intrinsic? It's explicitly referenced in AllocateArrayBuffer, but it doesn't appear in the table.

What about %AsyncFunction%? It's in the table, but isn't explicitly referenced by any algorithm. (Several implicit references though.)

It's possible that the first 'definition' isn't meant to be a definition, but is just a statement of something that is true of well-known intrinsics, but is also true of other intrinsics. In which case, it's not a very helpful statement.

Either way, this PR should change some prose in the Well-Known Intrinsics clause.

[jmdyck]

With this line removed, there's no explicit statement that the phrases "the Boolean Prototype Object" and "%Boolean.prototype%" denote the same object. While this equivalence might seem obvious given the similar wording, I don't think it should be left implicit.

Some possible solutions:

• Just reinstate the line.
• Leave the line out, and change:
  <p>The Boolean prototype object:</p>
  to something like:
  <p>The Boolean prototype object (i.e., %Boolean.prototype%):</p>
  or
  <p>%Boolean.prototype% (the Boolean prototype object):</p>
  (or maybe change the clause heading instead).
• Leave the line out, and replace all occurrences of "the Boolean Prototype Object" with "%Boolean.prototype%". (This would probably be clearer as a separate PR, and might also address the equivalence of "the Foo constructor" and "%Foo%", which is not touched by this PR.)

(Ditto all of the above for the other similar removals.)

[ljharb]

How is it implicit at all? It's the object on Boolean.prototype, Boolean is %Boolean%.

[jmdyck]

Sorry, when you say:

It's the object on Boolean.prototype

you mean that the phrase "the Boolean prototype object" denotes (the initial value of) Boolean.prototype, right?

I agree that that's the intent, but this PR doesn't say so. I'm not saying people won't figure it out, I'm saying there's no reason it should be something they need to figure out.

[ljharb]

I'm saying that %Boolean.prototype% is already in the spec as "the original value of the prototype property on %Boolean%" - so perhaps just the sentence here needs to be updated, since it's currently a tautology?

[jmdyck]

I'm saying that %Boolean.prototype% is already in the spec as "the original value of the prototype property on %Boolean%"

Oh, okay. Well, I agree that that's in the spec, but I don't see how that answers my original point.

• so perhaps just the sentence here needs to be updated, since it's currently a tautology?

I agree that's currently a tautology. How would you update it? Maybe just demote it to a Note?

[ljharb]

I don't think it needs even a note; just making sure that "the Boolean Prototype object" is also noted somewhere as "the value of .prototype on %Boolean% should suffice?

[jmdyck]

just making sure that "the Boolean Prototype object" is also noted somewhere as "the value of .prototype on %Boolean% should suffice?

Okay, I'm not sure how we got here, but yes, this aligns with two of the solutions I originally suggested.

And the third solution was to eliminate the phrase "the Boolean Prototype object". Does the spec benefit from introducing that phrase, rather than just saying "%Boolean.prototype%" (or Boolean.prototype, suitably qualified)?

[ljharb]

built-in objects that are explicitly referenced by the algorithms of this specification

To me this includes every object in the entire spec, which includes, eg, %ArrayBuffer.prototype%. I agree the prose needs a bit of tweaking here, though, to avoid the ambiguity.

[ExE-Boss]

I think we should keep the <dfn>s for %*.prototype% to keep auto‑linking working.

See also:

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

[ExE-Boss]

Suggested change

	<ul>
	<ul>
	<li>is the intrinsic object <dfn>%Object.prototype%</dfn>.</li>

[ExE-Boss]

When I posted #2056 (review), I thought that you’d add the <dfn>s for the remaining %*.prototype% properties without me needing to list them manually.

[Jack-Works]

does the change make undeniable intrinsics deniable?

[ljharb]

@Jack-Works this PR is purely editorial (no observable change) except for an impact on Function.prototype.toString (which makes this PR accidentally normative, which means it won't land soon). Deniability is unrelated.

[ExE-Boss]

Well, the impact on Function.prototype.toString won’t be observable, as engines already include PropertyName when stringifing all named built‑in functions.

Also, I’m pretty sure that dotted intrinsics also count as well‑known intrinsic objects.

[ljharb]

@ExE-Boss as is, the PR is normative in the sense that it allows engines to omit more names. Another option is to tighten the toString requirement to require all names.

"well-known" means it's in the table; I don't think dotted forms are in that category - and if they were, then the change to support the dotted notation was normative, because it then imposes the toString name requirement on every function.

[ExE-Boss]

We could also keep the dotted function intrinsics in the Well‑Known Intrinsic Objects table until #2068 is resolved.

[ljharb]

#1948 or similar would need to land before this is non-normative.

8/7 update: #1948 has landed; this is now editorial.

[devsnek]

Couldn't we technically just remove this whole table? Every intrinsic is already declared inline with the relevant definition via "is the %Xyz% intrinsic object"

[ljharb]

@devsnek i think it's useful for the table to exist for all the top-level intrinsics; but either way it'd be out of scope of this PR.

[jmdyck]

You eliminated this line, and the corresponding one for WeakSet.prototype, which makes sense because they're tautologies now. But why just these two? There are about 20 others.

[ljharb]

#2131 should address these; i'll rebase this after that lands and clean it up.

[jmdyck]

Looks okay to me.

[syg]

lgtm, long time in the making, congrats!