2020-10-01 [19:20:17.0000] have we suffered enough to allow a `has` syntax or smth yet [19:42:16.0000] devsnek: not yet, i fear. but i'd immediately champion such a proposal if it seemed like it had a chance of proceeding [23:21:43.0000] in spec.emu, If I need to add a new section under an existing section, how do I go about it? [23:26:45.0000] just... add a section? [23:26:56.0000] I maybe do not understand the question [23:27:41.0000] you just add ` [your stuff] ` wherever it's supposed to go [01:13:52.0000] Bakkot: thanks, was lost in the nesting. [11:37:03.0000] i'm sobbing https://bugzilla.mozilla.org/show_bug.cgi?id=1668342 [11:37:12.0000] proposal to add [[IsHTMLDDA]] to .item methods [11:42:09.0000] lol yeah [11:42:23.0000] since that's paid-for code, hopefully that can be fixed [11:42:46.0000] [[IsHTMLDDAOrItemMethod]] [11:42:59.0000] i hope those sites can be updated lol [11:43:26.0000] even if magic360 patches their library, sites might be disconnected from that [11:45:35.0000] I do not have much hope of it [11:47:44.0000] 1. The production `E.length && E.item` must evaluate to *false* when `E` is an Array exotic object. [11:49:23.0000] oh no [12:24:27.0000] /me cries softly in webcompat [12:25:37.0000] but like, given that this proposal was motivated by layering, I thought it was understood that there would be "upward effects" no matter what? [12:26:02.0000] still, that line is worrying [12:30:44.0000] the hope was that things wouldn't break [12:34:57.0000] very true :( [12:35:18.0000] maybe that's *your* hope [12:35:24.0000] you gotta enjoy the chaos [12:40:36.0000] I do not have to enjoy the chaos [12:40:42.0000] I have had quite my fill of chaos this year [12:41:05.0000] though I gotta admit I am a little less invested in the proposal's success when it has String.prototype.item riding with t [12:42:11.0000] I just am hoping not to have to revert it [12:45:24.0000] with this particular problem, i think string item is the only survivor [12:45:30.0000] ironically enough [12:45:33.0000] devsnek: where is that line from? I can't google it [12:45:44.0000] ljharb: strings have length though [12:45:48.0000] rkirsling: my head? idk [12:46:11.0000] oh what?! I thought that was in a WHATWG spec [12:46:21.0000] "you gotta enjoy the chaos"? [12:46:30.0000] no no the spec line [12:46:41.0000] oh lmao [12:46:46.0000] 😅 [12:46:53.0000] sorry, I was unclear [12:46:56.0000] `E.length && E.item` [12:47:03.0000] `E.length && E.item` is the line of code that is in that magic360 spinner thing [12:48:11.0000] right. but you weren't saying that that's an invariant of some spec? [12:49:26.0000] rkirsling: i was jokingly saying we should add it to ecma262 [12:49:33.0000] ohh phew [12:49:50.0000] you got me hook line and sinker [12:53:35.0000] devsnek: sure but the implication is that code is only checking objects. i certainly haven't dug into it [12:57:52.0000] i have so many places i want to use `.item(-1)` [12:58:12.0000] samesies [13:10:56.0000] does anyone have opinions about this https://github.com/tc39/ecma262/pull/2125#discussion_r497861480 [13:16:10.0000] I would prefer we not do that approach [13:16:32.0000] the "a |Script| which is being evaluated for JSON.parse" prose is much, much clearer [13:17:17.0000] feels too magical for my tastes [13:17:47.0000] open to other approaches, but clarity is a much higher priority for me than lack of magic [13:17:58.0000] this is a document for humans, not a computer program [13:19:48.0000] hmm we need coverage for `JSON.parse('{ "__proto__": x, "__proto__": y }')` [13:32:05.0000] is this a ? 🦋 `#{ length: 0, item: true }` [13:32:30.0000] gibson042: no [13:32:40.0000] but `#{ length: 1, item: true }` is [13:33:05.0000] oh man, please meme-ify and tweet that :D [13:34:45.0000] I think it's probably poor form for committee members to tweet memes making fun of code from random web devs [13:36:23.0000] oops [13:36:37.0000] that's true 😓 [15:52:35.0000] yes let’s please not disparage the magic 360 library. both firefox and chrome have reached out, let’s see how it goes [15:54:24.0000] apologies [15:54:31.0000] I hope the communication goes well [15:58:43.0000] i mean it’s fine to vent, but we shouldn’t insult them as a way to get them to update [15:59:04.0000] agreed 2020-10-02 [17:35:17.0000] If the magic 360 library is a paid library, does it possible to contact the team? Because users have to pay to use the library so the team must have a list of their clients [18:29:32.0000] jackworks: Mozilla and Google have reached out [18:45:34.0000] Cool [21:44:28.0000] rickbutton: btw the current logic in the spec text considers `#[0]` and `#[-0]` equal but not `#[#[0]]` and `#[#[-0]]` [22:05:07.0000] are you sure? [22:05:07.0000] SameValueZero -> TupleSameValueZero [22:05:07.0000] for each element: [22:05:07.0000] SameValueZero -> TupleSameValueZero [22:05:18.0000] reads as correct to me, but also it is late [22:06:15.0000] rickbutton: nvm there was a typo :P sorry [22:06:25.0000] u good :) [22:06:43.0000] is the Tuple(len) override for compat with Array(len)? [22:06:53.0000] it seems like something worth getting rid of [22:10:20.0000] yeah it was originally for symmetry, we have a PR to get rid of it: https://github.com/tc39/proposal-record-tuple/pull/173 [22:10:20.0000] we haven't made a final decision but I don't think anyone on the champion group has a strong opinion either way [22:12:23.0000] the Array ctor overload s generally considered to have been a mistake [22:12:30.0000] I am in favor of not repeating it [22:12:37.0000] slightly more of a mistake than computers themselves [22:13:13.0000] alternatively, do not allow it to be invoked at all, and instead only allow .of and .from [22:13:17.0000] (this is probably a bad idea) [22:13:49.0000] this has made the very sparse wall of Things I Lack an Opinion On [22:14:00.0000] i don't mind very much what Tuple(x) does [22:14:12.0000] as long as it doesn't magically create a tuple of [x] or len x based on the type of x [22:46:36.0000] how come https://tc39.es/ecma262/#sec-array.prototype-@@unscopables includes, e.g., flatMap but not map? [22:47:01.0000] rkirsling because `map` already existed when unscopables was introduced [22:47:12.0000] hm [22:47:15.0000] so someone using `with (arr)` presumably had already accounted for `map` [22:47:17.0000] pretty sure only ES6+ things are unscopables [22:47:39.0000] but presumably had not accounted for `flatMap`, since it didn't exist [22:47:50.0000] my real question: are we supposed to add `item` then? [22:47:53.0000] yes [22:47:55.0000] yes [22:47:57.0000] k [22:48:08.0000] any new string-named array prototype property [22:48:23.0000] like it says in the note: The own property names of this object are property names that were not included as standard properties of Array.prototype prior to the ECMAScript 2015 specification [22:48:30.0000] should TypedArray.prototype have it tho? [22:48:41.0000] TA should not have unscopables [22:48:42.0000] does it? [22:48:51.0000] just Array currently [22:48:56.0000] good good [22:49:20.0000] (TA should, and does, get `.item`, if that was the question) [22:49:44.0000] it doesn't have it [22:49:46.0000] i'm wondering why it doesn't [22:50:01.0000] TA didn't exist prior to ES2015 [22:50:06.0000] there was no back compat to worry about [22:50:21.0000] unscopables is purely a back compat thing, since new code should not have `with` [22:51:04.0000] with considers non-enumerables and non-owns too tho, so why doesn't RegExp, Date, etc have unscopables? [22:51:36.0000] presumably there was not a concern that anyone was writing `with (regex)` [22:51:46.0000] or `with (date)`, etc [22:52:04.0000] actually I don't know if Date or RegEx got any new string-keyed methods [22:52:15.0000] i don't think they did [22:52:15.0000] RegExp got a ton of new prototype accessors [22:52:21.0000] but i suppose they were owns before [22:52:29.0000] so, .flags, .unicode, .sticky [22:52:40.0000] String got .matchAll, .includes [22:52:53.0000] ah, yeah [22:53:08.0000] so yeah, presumably there was not concern that people were `with`-ing such things [22:53:09.0000] also wouldn't we want to actively prevent someone from doing something bad with `with (promise)` or `with (map)` etc? [22:53:15.0000] certainly that would be an odd thing to do [22:53:18.0000] right but like, what's the added cost then to including all the unscopables [22:53:32.0000] its assumed that code doing `with (array) {}` exists [22:53:32.0000] a bunch of overhead in engines? [22:53:41.0000] it would only slow down code that did the odd things that nobody thinks will ever happen [22:53:48.0000] and adding new methods shouldn't add new variables into those scopes [22:53:57.0000] no, it adds to the amount of stuff you need to load into memory when you load the page [22:54:02.0000] it's not free [22:54:22.0000] don't modern engines put everything in .ro_data anyway [22:54:39.0000] until it gets messed with anyway [22:54:40.0000] Bakkot: i mean, it could be lazy-loaded only when accessed or used with `with`, and if that never happened then, free? [22:54:57.0000] re: "wouldn't we want to actively prevent someone from doing something bad with `with (promise)` or `with (map)` etc" - no? we already have a sensible thing, which is, strict mode prevents you from doing `with`; I don't think we really care to round off sharp edges for people who do that [22:55:10.0000] I don't think lazy-loading was nearly as common a technique when es2015 was introduced [22:55:15.0000] how is array prototype unscopables not rounding off ♯ edges [22:55:21.0000] back compat [22:55:31.0000] i think array prototype ones is about not breaking code [22:55:34.0000] it's not "you might doo something bad", it's "we might break the web" [22:55:37.0000] very different issue [22:55:37.0000] ah k [22:55:53.0000] `with (array) { flatMap(array, x) }` [22:56:02.0000] so, we could add them to all the others, and it'd be a cleaner system, and it wouldn't break anyone, but nobody cares because nobody does that? [22:56:18.0000] sounds right to me [22:56:27.0000] i dunno if i would say cleaner [22:56:39.0000] well, yeah, "cleaner" is disputable [22:56:47.0000] already crazy enough that unscopables is an object [22:57:19.0000] sure, fair [22:57:45.0000] yeah it's super weird it wasn't just a Set [22:58:51.0000] eh, object is a reasonable enough choice if you only want string keys [22:59:32.0000] sets get weird - like, does it reach into the slot directly, making Set contents unforgeable? does it invoke the iteration protocol, with all the complexity that entails? [22:59:48.0000] hm, true [23:00:53.0000] I use Object.create(null) objects rather than sets or maps in some of my performance-sensitive code for that reason, when they suffice [23:00:54.0000] should just be array#includes [23:01:18.0000] includes performs observable property accesses [23:01:26.0000] indeed [23:01:30.0000] oof [23:01:38.0000] i guess back in 2015 engines didn't really have the optimization layout for making that ok [23:01:41.0000] but today they do [23:03:31.0000] making use of that here sounds like it would add an extremely nasty code path to an already extremely nasty part of engines [23:04:42.0000] should've just disabled with in strict and sloppy [23:04:53.0000] sure the realms polyfill couldn't exist, but i see that as a bonus [23:07:40.0000] and `eval` while they were at it, presumably [23:08:11.0000] I used eval in anger this week; it was fun [23:08:23.0000] makes prototyping a code generator a lot easier [23:08:47.0000] lol [23:08:52.0000] just mash some strings together and eval it; don't even need to write to disk [23:09:15.0000] i use the vm to get rid of the guilt [23:09:17.0000] gonna switch to constructing and realizing ASTs at some point [23:09:19.0000] vm api* [23:09:34.0000] probably around the time I switch to doing the code generation in Java [23:09:40.0000] ohno [23:09:49.0000] (the week before that I wrote eval for java! good times, good times.) [23:10:18.0000] (don't worry, it's only used in tests) [23:10:24.0000] O.o [23:11:09.0000] would not want to work on/in Java [23:11:16.0000] it's actually pretty nice! [23:11:35.0000] must be subjective [23:11:56.0000] i've written java professionally for a few months total, and i doubt anyone would pay me enough to do it again [23:11:58.0000] has some painful bits, but version 8 and subsequent introduced a lot of nicer parts [23:12:00.0000] we used java in my cs classes [23:12:11.0000] never going back [23:12:54.0000] I think java-as-taught-in-cs-class is probably a much worse dialect of the language [23:13:33.0000] at least i skipped the class where java itself was the subject [23:14:58.0000] probably the worst thing was failing an assignment because the homework checker didn't support `var` [23:15:44.0000] heh https://gc.gy/69324339.png [23:16:12.0000] java has lambdas! [23:16:25.0000] ya but you have to declare an interface [23:16:39.0000] that type is normally spelled `Predicate` [23:16:57.0000] I believe it's built in [23:17:00.0000] shouldn't need to declare it [23:17:01.0000] neat [23:17:18.0000] https://docs.oracle.com/javase/8/docs/api/java/util/function/Predicate.html [23:17:20.0000] i'll remember that if i ever write java ever again :P [23:17:37.0000] heh 2020-10-03 [00:04:57.0000] `${Number.prototype}` => '0', but why? [00:05:15.0000] howdoi: historically, all the builtin prototypes were exotic, and an instance of themselves [00:05:32.0000] howdoi: in ES6, many of them were changed to be ordinary, but some couldn't be for web compat. number, function, regex, date, and a few others probably [00:07:57.0000] ljharb: ha, so there are few checks for internal slots instead of the primitive directly, for the same reason? [00:14:02.0000] howdoi: not sure what you're asking [00:19:01.0000] ljharb: https://tc39.es/ecma262/#sec-properties-of-the-number-prototype-object 2.a [00:19:28.0000] howdoi: those are boxed primitives [00:19:32.0000] check ToObject() [00:22:00.0000] howdoi: right, every `thisNumberValue` either returns the number or unwraps the boxed number, or throws. [00:23:30.0000] I guess the question is - does the spec have a easy way to check instanceOf within its steps? If yes, then why is it (sort of) duck-type checking about Number instances (using type object and [[NumberDatta]] slot) ? [00:24:13.0000] Is it for some particular case? Or just the easiest way to check instanceOf Number? [00:24:20.0000] devsnek: nods [00:25:00.0000] bendtherules: `thisNumberValue` works no matter what realm [00:25:16.0000] instanceof is limited to checks within a realm [00:25:39.0000] makes sense [00:25:50.0000] If Type(value) is Object and value has a [[NumberData]] internal slot, then there is a check for value.[[NumberData]] [00:26:51.0000] like ljharb said, Assert: Type(n) is Number, throw if not. [00:30:42.0000] /me Number {value,type,object} hmm... [00:33:41.0000] thanks ljharb && devsnek [00:34:25.0000] bendtherules: slots tell you what the thing is. instanceof is checking something else, something that’s often useless. [00:36:02.0000] Ok, so how does spec check if something is a array? [00:36:29.0000] IsArray [00:36:51.0000] probably the single most cursed operation in the spec lol [00:37:28.0000] lol [01:31:05.0000] JSC, SM, V8, and also babel, acorn, esprima all accept `(async () => await 1 ** 1)()` [01:31:16.0000] but that's not actually legal, right? [01:32:22.0000] also engine262 [01:32:37.0000] and shift [01:32:48.0000] not xs or chakracore though! [01:37:40.0000] Bakkot: did we all forget that await expressions are unary expressions? [01:37:47.0000] apparently! [01:37:52.0000] everyone except typescript [01:38:05.0000] (came up in https://github.com/microsoft/TypeScript/issues/40916 ) [01:38:30.0000] where even is the early error for ** [01:38:32.0000] i can't find it [01:38:38.0000] still possible there isn't one [01:38:44.0000] it simply doesn't match the grammar [01:38:56.0000] er, sorry, "still possible" was the start of a different sentece [01:39:04.0000] there is no early error, it simply doesn't match the grammar [01:39:24.0000] and, separately, it's still possible I'm misreading the spec and `await 1 ** 1` is legal [01:39:28.0000] hm i have it as an explicit check in parsing a unary expression [01:39:59.0000] hmm [01:40:01.0000] oh i see [01:40:07.0000] very interesting [01:40:31.0000] ExponentiationExpression : UnaryExpression [01:40:34.0000] we need to put a big note on that [01:40:51.0000] i think everyone scans over that assuming everything there is just straight precedence climbing [01:41:53.0000] well, people get the -1**1 bit right [01:42:03.0000] presumably there's coverage though [01:42:08.0000] and also it's well known [01:43:35.0000] is there any reason AwaitExpression is separated into its own production [01:43:44.0000] could just be `[+Await] await UnaryExpression` couldn't it? [01:44:15.0000] hm i guess it makes `contains AwaitExpression` cleaner [01:44:19.0000] yeah, that [01:46:01.0000] i can submit a fix to v8 i suppose [01:46:09.0000] i hope this isn't web reality now [01:47:26.0000] it might well be [01:47:40.0000] `await x ** 2` is a reasonable thing to write [01:48:44.0000] all the keyword ones are reasonable [01:48:56.0000] delete/void/typeof/await [01:49:01.0000] urgh, I hate that that is how it's spec'ed :( [01:49:19.0000] those are not reasonable [01:49:37.0000] you would not want to exponentiate the result of delete or void or typeoof [01:49:52.0000] as reasonable as `await x ** 2` anyway [01:50:49.0000] `await x` can reasonably result in a number, which is a thing you might want to square [01:51:00.0000] it really is different from the others here [01:51:59.0000] but no one would read it as `(await x) ** 2` [01:52:48.0000] ehhh, depends on how familiar with the rules you are [01:52:59.0000] i mean that argument flies for `-1 ** 1` [01:53:28.0000] well, right, the point of the `-1 ** 1` thing is that both interpretations were plausible [01:54:00.0000] my claim is not that the `(await x) ** 2` interpretation is obviously right, just that it's plausible enough that people might have written it in real code [01:54:15.0000] ah i see [02:05:07.0000] `await x ** 2` seems way more harmful than `-1 ** 1` to my eyes which did not expect to encounter this [02:05:22.0000] filed https://github.com/tc39/ecma262/issues/2197 [02:05:44.0000] /me hopes there is web reality [02:05:46.0000] rkirsling: fwiw `await a + b` is definitely legal [02:05:56.0000] oh [02:06:01.0000] hm [02:06:04.0000] fair enough [02:06:21.0000] and everybody does the right thing, I assume? [02:06:32.0000] haven't checked [02:06:34.0000] presumably [02:06:53.0000] if so then I rescind my whining [02:07:32.0000] this is one of those places prettier will insert parentheses for you [02:07:58.0000] (it does for `await a ** b` as well, funnily enough) [02:09:37.0000] yeah with a linter there's nothing to worry about it [02:09:44.0000] er formatter [09:20:45.0000] https://www.python.org/dev/peps/pep-0638/ [09:21:10.0000] bold of a PEP to contain the phrase "compile-time" [09:22:03.0000] i could live with js never having macros [09:22:08.0000] its dynamic enough to not need them [09:23:01.0000] js has macros; they' [09:23:06.0000] re called "babel" [09:24:06.0000] lol [09:24:25.0000] mfw engine262 uses babel for macros [09:33:22.0000] > Currently, all AST nodes are allocated using an arena allocator. Changing to use the standard allocator might slow compilation down a little, but has advantages in terms of maintenance, as much code can be deleted. [09:33:37.0000] oh to be a cpython dev 2020-10-04 [13:21:07.0000] https://twitter.com/jennschiffer/status/1312825920454041608 [13:21:19.0000] people want a larger standard library and types [13:21:44.0000] and one guy wants more consistency between pass by value and pass by reference [13:39:23.0000] devsnek: just replied to that last guy, JS is only and always pass by value :-) [13:39:36.0000] lol [14:24:59.0000] those terms are not super useful [14:26:42.0000] that's also true [14:27:25.0000] not quite as bad a "strongly typed" and "weakly typed", but kinda in the neighborhood [14:35:44.0000] pass by handle 2020-10-05 [20:12:28.0000] anyone know how to hit step 8.c of AsyncGeneratorYield? [20:12:41.0000] i thought it would be like `it.return(Promise.reject())` but that doesn't seem to do it [21:46:06.0000] devsnek: make a Get of `constructor` on an awaited promise throw? [21:46:20.0000] huh [21:46:58.0000] like `return({ get constructor() { throw new Error() } })`? [21:48:24.0000] yeah maybe? not sure [21:48:34.0000] the only operation that has a ? in AsyncGeneratorYield is PromiseResolve [21:49:07.0000] which has a ? on `Get(x, "constructor")` and also on constructing the constructor [08:10:20.0000] i remain surprised that people are just _so_ opinionated about programming languages [08:10:35.0000] (re: jenn schiffer's tweet) [08:10:46.0000] shu: is that sarcasm [08:11:01.0000] no [08:11:29.0000] people in tc39 are pretty opinionated [08:11:48.0000] out of things in computing to devote opinion budget to, i wouldn't put PLs so high personally [08:45:47.0000] interesting thoughts at https://twitter.com/HarryB/status/1312860441639444480 [08:47:45.0000] i'm down with everything being expressions [08:47:47.0000] aside from declarations [08:50:46.0000] i'm not sure how versioning solves the problem of catering to a common denominator [09:05:24.0000] versioning is like regexps, now you have n problems [09:17:21.0000] O(2^n) problems [09:18:16.0000] yeah, after I hit enter I was like, it's definitely more than n [09:44:19.0000] that's why I asked for clarification... it may be that e.g. 80% of versioning pain can be addressed with simple solutions like straightforward capability detection and/or conditional loading [13:06:17.0000] I'm p sure the "versioning" thing is "I want to serve script X to this browser, but script Y to other browser" [13:21:24.0000] it was actually super awesome when we had IE conditional comments, but the landscape is way more complex than that now [13:58:07.0000] Sigh, yeah, those were great. [15:19:38.0000] can somebody explain how this test corresponds to the spec? https://test262.report/browse/built-ins/TypedArrayConstructors/from/set-value-abrupt-completion.js [15:19:50.0000] because 6.a @ https://tc39.es/ecma262/#sec-%typedarray%.from calls IterableToList [15:20:01.0000] and then we use the produced list [15:21:54.0000] so that should imply that even if you get an element that throws from `valueOf`, you've still finished iteration of the source [15:21:55.0000] (i.e. it's specifically `assert.sameValue(lastValue, obj, "interrupted source iteration");` that isn't making sense to me) [15:25:34.0000] rkirsling: hm, the only thing i remember around that is https://gist.github.com/ljharb/896ad592accdbd783d5ec1d44e978b76 that i presented in july 2019 [15:25:48.0000] i didn't look at interrupted iteration at all [15:28:43.0000] hmm, there isn't a corresponding test for Array.from but then the Array constructor doesn't require a size param [15:28:55.0000] are there any blog posts about how jsc regex is so fast [15:32:48.0000] there is a section of one, in here: https://webkit.org/blog/8685/introducing-the-jetstream-2-benchmark-suite/ [15:33:44.0000] wonder if I should make a GH issue for my question [15:40:10.0000] oh wait I get it [15:41:27.0000] it's not a question of last-iterated, it's a question of last-mapped [16:10:51.0000] ljharb: this was what threw me, in case you were curious: https://bugs.webkit.org/show_bug.cgi?id=217349 2020-10-06 [18:23:06.0000] ljharb: I can squash 2176 overloaded-functions when/if the time comes. 2020-10-07 [18:26:27.0000] Greetings [09:37:04.0000] who is in charge of landing test262 changes? [09:37:16.0000] rick? [10:09:06.0000] he's done all the merges since Sept 9, so it seems like a good bet. [12:19:18.0000] oof https://github.com/tc39/proposal-item-method/issues/31 [12:33:10.0000] oof indeed [12:35:48.0000] obvious solution: stop having `String.prototype.item` [12:35:52.0000] it is bad anyway [12:35:54.0000] :D [12:37:40.0000] there's the other one anyways that might block Array.prototype.item [12:38:19.0000] yeah I am more sad about the actually good one being web-incompat [12:38:23.0000] I mean thankfully the two are separate but [12:38:35.0000] I really hope we can resolve the Array one [12:39:27.0000] also GH's reaction set is not "right-sized". I don't want to react with 😕 I want to react with :sob [12:39:31.0000] 😭 even 2020-10-08 [19:45:58.0000] yeah i am swayed by that to remove String#item [19:46:15.0000] i have also reached out to Magic 360 again since they have yet to respond [19:47:28.0000] it is my wish that when i do bring this up again [19:47:43.0000] that we refrain from talking more about code units [19:47:54.0000] and let the web compat argument stand alone and leave it at that [19:49:07.0000] couldn't agree more [20:03:07.0000] I will avoid bringing it up [20:18:01.0000] some thoughts about growing C++'s stdlib, many of which apply to us as well: https://cor3ntin.github.io/posts/std/ [22:38:53.0000] shu: if magic360 and their customers can't be updated, what happens? [22:39:44.0000] weeping and gnashing of teeth [22:40:13.0000] like, i assume that the original motivation is gone, but that there's still a value in having the semantics somehow [22:41:36.0000] it's hard to say; the champions said they would abandon the proposal if `item` were unviable [07:43:09.0000] ljharb: what do you mean the original motivation is gone? [07:46:50.0000] shu good morning! Do you have a moment to weigh in on this: https://github.com/tc39/test262/pull/2833#discussion_r500652041 ? [07:47:48.0000] rwaldron-: sure, ross's understanding there matches mine and he did confirm with me yesterday on irc [07:49:00.0000] shu great, then I believe these tests are ready, in support of the spec change. [07:50:34.0000] great, thank you [07:52:24.0000] Oh, I'm sorry: I didn't even see that rkirsling had edited his last comment. [09:18:34.0000] gotta say the YUI thing doesn't look great [09:19:32.0000] shu: i meant that wasn’t the motivation to make everything match observablearray? [09:19:52.0000] ie, item or nothing? [09:21:10.0000] So what's the next step? Give up the item proposal? [09:21:14.0000] ljharb: that was the motivation to make it named `item` [09:21:25.0000] ljharb: the semantics already doesn't match IDL's current semantics [09:21:28.0000] shu: ah ok cool, so it’s still a thing with a different name? [09:21:33.0000] if only we had a language-level brand checking mechanism the web could adopt, people would stop ducktyping builtins :-/ [09:22:32.0000] that'd only work for newer things [09:23:10.0000] also people would not stop ducktying builtins [09:23:10.0000] I don't think we need "item" it cannot resolve the DOM observable array. [09:23:11.0000] newer code, sure, but older code might migrate over time [09:23:13.0000] jackworks: good question. i'm not sure right now. Magic 360 got back to me that they did update and will encourage their users to update [09:23:23.0000] there's plenty of other ways to brand check [09:23:28.0000] I mean "If it cannot" [09:23:34.0000] ljharb: if "older code might migrate over time" was our hope, we'd operate a lot differently [09:23:35.0000] Bakkot: not elements or node lists [09:23:45.0000] there is, you can do `.constructor.name` at leat [09:23:46.0000] least [09:23:49.0000] shu: yeah fair. Encouraging news from magic360 [09:24:00.0000] that’s pretty unreliable [09:24:07.0000]