2016-09-01 [22:47:53.0000] annevk: yt? [22:53:20.0000] rniwa: yt? [22:53:25.0000] dominicc: yeah [22:53:35.0000] dominicc: so there's some confusion about brand check in https://dom.spec.whatwg.org/#concept-create-element [22:54:01.0000] dominicc: I think the intent is to check it against any HTMLElement interface, not just the one in the same global [22:54:43.0000] Can I try custom elements in WebKit Nightly? I ran your LayoutTests in Chrome yesterday, they were very informative, found a few good bugs. I also found some things which, if WebKit is passing the tests, might be contrary to the spec. I made a detailed list I can share with you if you're interested. [22:55:40.0000] dominicc: great :) [22:55:41.0000] dominicc: that would be graet [22:55:41.0000] great* [22:55:41.0000] dominicc: unfortunately, there is not a good way to enable a runtime flag at the moment :( [22:55:41.0000] I have a link, one second... [22:55:41.0000] dominicc: you basically have to modify the code :( [22:55:50.0000] I tried building WebKit but I think I need a newer framework; I have 10.10. Guess that is old. Got some macro failure spew about CF_AVAILABLE. [22:56:14.0000] dominicc: hm... [22:56:21.0000] dominicc: so I'm going to try enabling it in STP soon-ish [22:56:30.0000] dominicc: my goal is to finish implementing as much as I can before TPAC [22:56:36.0000] dominicc: so that we can just sit down & compare impls [22:56:46.0000] Here's my notes: http://goo.gl/yli9Ao [22:57:40.0000] Great! I plan to be at TPAC every day. [22:59:09.0000] re: confusion about the brand check, I guess it takes digging throw WebIDL to find out what "implements" means, but I could imagine that means any HTMLElement interface. What do you think? [22:59:41.0000] rniwa: I assume you mean in step 6.1.3? [22:59:48.0000] dominicc: yeah [23:00:35.0000] rniwa: There's a webidl issue linked in the note, it seems that any HTMLElement is certainly the intent. [23:01:02.0000] dominicc: yeah [23:02:58.0000] dominicc: cool. thanks for the notes [23:03:15.0000] dominicc: they were written way back in April or something when I first implemented the new spec [23:03:31.0000] dominicc: so all these things are fallouts from the spec changes that have happened since then [23:04:44.0000] dominicc: e.g. we don't implement any of sanity checks in 6.1.: https://dom.spec.whatwg.org/#concept-create-element [23:05:00.0000] although I'm fixing that right now. sort of. [23:16:42.0000] rniwa: around a bit, lack of branding is a known issue with IDL and DOM [23:17:14.0000] annevk, dominicc: for now, WebKit & Chrome implementations both do "the right thing". [23:17:20.0000] but this is a compat risk. [23:21:39.0000] rniwa: I guess we could clarify the issue and planned fix more, file an issue? [23:21:40.0000] annevk: I was about to but it seems like your issue is still open [23:21:40.0000] annevk: so i figured i don't need another one [23:21:40.0000] annevk: I'll just wait 'til the definition is added and then we can argue whether we need to fix it or not [23:21:40.0000] rniwa: IDL is on track to get maintained a bit more, but it'll take a while still I suspect [23:21:40.0000] rniwa: if Apple could find an editor somewhere… [23:21:40.0000] annevk: well, we need to first implement all these specs people have produced [23:21:40.0000] annevk: but yeah, i wish we had more man power to contribute there. [23:21:45.0000] rniwa: if only Apple had the money [23:22:14.0000] annevk: the money isn't usually issue though. finding a candidate is insanely hard :( [23:22:30.0000] annevk: we've had a position open for more than 12 months in our team for example [23:22:52.0000] heh [23:24:36.0000] annevk: the requirement that one has to work in Cupertino isn't helping there :( [23:25:16.0000] why does the location matter? [23:25:21.0000] annevk: so we can't tap on those good european software engineers who aren't willing to relocate, etc... :( [23:25:27.0000] boogyman: i don't know. [23:26:20.0000] maybe the fact that the position has been unoccupied for so long would spur the recruiters to consider other candidates [23:31:44.0000] annevk: are we supposed to be creating Element in a XML document when the custom element interface isn't defined and its name is a valid custom element name? [23:31:52.0000] annevk: or are we still supposed to create HTMLElement [23:37:15.0000] dominicc: ^ [23:37:23.0000] I think the current spec indicates that it should be Element. [23:37:31.0000] implies* [23:38:18.0000] rniwa: What does the HTML spec say about XML documents having browsing contexts? I think if you found a definition you'd create a HTML element; what do you think about the possibility of finding a definition? [23:38:54.0000] dominicc: I'm not following. [23:39:06.0000] rniwa: not sure [23:39:21.0000] dominicc: the existence of browsing context is independent of whether it's a XML document or a HTML docuemnt. [23:39:22.0000] To create a custom element you have to find a definition; to find a definition, you have to be creating an element in the HTML namespace. [23:39:25.0000] rniwa: forgot whether XML parser is patched [23:39:41.0000] dominicc: that's not entirely true. [23:39:46.0000] That is https://html.spec.whatwg.org/multipage/scripting.html#look-up-a-custom-element-definition step 1? [23:40:06.0000] dominicc: you can create an element inside a XML document, which uses the same global object as the one that as a browsing context [23:40:17.0000] dominicc: and then adopt that element into the document with a browsing context [23:40:18.0000] rniwa: Are you referring to create an element step 6.2.1? [23:40:30.0000] dominicc: not sure [23:40:34.0000] dominicc: not quite* [23:40:54.0000] dominicc: I'm trying to create an element in a detached XML document. e.g. one created by DOMImplementation [23:41:22.0000] dominicc: then adopt it back into a HTML document with a browsing context both of which shares a single global object [23:41:33.0000] /me this is getting complicated [23:41:59.0000] rniwa: OK. I'm with you so far. You do those things. [23:42:01.0000] dominicc: so in this case, the element we're creating isn't necessary in HTML namespace. [23:42:22.0000] dominicc: so I think it should be using Element interface [23:42:37.0000] dominicc: and upgrade should subsequently fail because it doesn't match the namespace [23:43:06.0000] more precisely, looking up a custom element definition step should not find any interface [23:43:13.0000] rniwa: Right. I think every algorithm checks the namespace is HTML and bails out/no ops if it isn't HTML. [23:43:19.0000] dominicc: something like that [23:43:31.0000] Right. That is look-up-a-custom-element-definition step 1 I linked to ^^^ [23:45:15.0000] dominicc: ah, okay. now I follow what you were saying earlier [23:45:39.0000] rniwa: Sorry for the bit about browsing contexts, you're right it's not a browsing context thing. [23:48:49.0000] dominicc: oh, np. [00:18:54.0000] dominicc: i forgot to reply but i'll be at TPAC for five days so hopefully we'll have a plenty of time to sort out details of custom elements [00:19:11.0000] rniwa: splendid, look forward to it! [00:19:24.0000] rniwa: domenic is coming too. [00:19:29.0000] dominicc: cool [00:20:08.0000] dominicc: luckily, custom elements have fewer contentious points than shadow DOM still [00:20:41.0000] dominicc: except is= [00:21:05.0000] dominicc: so hopefully it'd be straight forward to sort out the remaining issues at TPAC at least for autonomous custom elements [00:21:12.0000] rniwa: Yep! [00:22:09.0000] rniwa: We're working hard to get our tests in shape by that time. We need to do some updates for recent spec changes. [00:22:36.0000] dominicc: yeah, i'm doing the same for webkit tests as fast as I can [00:25:22.0000] dominicc: it looks like you haven't added "element definition is running flag"? [00:25:54.0000] rniwa: not yet. We still have the older check where reentrancy is allowed, as long as it is a different name. [00:26:05.0000] dominicc: I see. [00:26:06.0000] rniwa: This will change next week. [00:29:28.0000] dominicc: here's a way to enable custom elements on STP. [00:29:41.0000] rniwa: ooh goody, what? [00:29:42.0000] dominicc: on macOS Sierra, there's a submenu in "Develop" menu [00:30:22.0000] rniwa: interesting. I will have to set up Sierra... I'm travelling next week :/ [00:30:34.0000] dominicc: yeah, probably not a good idea :( [00:32:52.0000] dominicc: but going forward, you should be able to toggle experimental features on that submenu. [00:33:17.0000] so hopefully this pain of not being able to toggle features without having to build webkit will go away soon [00:50:15.0000] I wonder if we can start checking the MIME type for (classic) Worker()/SharedWorker() and importScripts() [00:51:44.0000] although maybe that doesn't help anything if [05:25:00.0000] or at least the effect is that document.characterSet in that case is windows-1252 [05:26:45.0000] hmm i suppose i should write tests in https://github.com/html5lib/html5lib-tests/tree/master/encoding [05:27:05.0000] are those in wpt also? [05:28:13.0000] no, partly because they're a total mess :\ [05:28:46.0000] we really need to deal with having tests for what the pre-scan is expected to find for a given pre-scan length, and what the eventual encoding at the end of parsing is [05:31:40.0000] gsnedders: so what should i do if i want to write new tests for https://github.com/whatwg/html/issues/1438 ? [05:32:52.0000] a) cry. b) ??? c) write tests. [05:33:09.0000] i could write straight wpt tests [05:35:24.0000] though it seems attractive to write in html5lib-test format and have a script generate wpt files [06:24:31.0000] annevk: are you doing anything fetch specific at TPAC? If not, happy to slant the service worker meeting towards fetch issues that are vaguely SW related [06:27:22.0000] MikeSmith: payments meeting not on today? [06:27:43.0000] (I realise I missed the first half) [06:54:30.0000] JakeA: was thinking we should have a session at some point [06:54:50.0000] JakeA: nothing organized though, so sounds good [08:56:25.0000] Hi! [08:56:41.0000] I would like to work in this issues https://github.com/whatwg/html/issues/507 [08:56:53.0000] my github account marti1125 [09:42:43.0000] marti_: you should be able to work on it by creating a fork [09:43:06.0000] marti_: and then creating a PR with your commits that implement the required changes [09:43:49.0000] marti_: if you're not familiar with pull requests, I found this that will probably help you out: https://yangsu.github.io/pull-request-tutorial/ [09:44:22.0000] (also, hi!) [09:46:41.0000] thanks :D [09:47:33.0000] for Remove top-level ForceType text/html is only remove this line https://github.com/whatwg/html/blob/master/.htaccess#L3 [09:47:43.0000] or remove all ForceType text/html [09:52:14.0000] marti_: just that line : [09:52:18.0000] :) [09:54:17.0000] ok thanks!! [09:55:02.0000] I will send a PR from my fork master to master [09:55:10.0000] or I have to create a new branch=? [09:55:39.0000] marti_: either will work, but creating a new branch will be easier for you if you want to work on other things in the future. Then you can do one branch per thing. [14:09:51.0000] It's here!! https://heycam.github.io/webidl/ \o/ yay tobie [14:10:06.0000] :) [14:10:26.0000] \o/ [14:12:23.0000] mkwst: ping [14:13:57.0000] Domenic: favorite glitch so far: Function (https://heycam.github.io/webidl/#Function) randomly referencing a Houdini spec. [14:14:17.0000] nice :P [14:18:01.0000] /me thought "houdini spec" was a joke [14:18:10.0000] apparently it's real [14:32:36.0000] KiChjang: well, I certainly wish it would magically disappear from the references on the spec. :D [14:35:46.0000] tobie: I'm sending the Houdini spec a change to avoid re-defining Function. [14:36:18.0000] jyasskin: oh, is _that_ why it's happening? [14:36:28.0000] Do I shave the yak that URL has url instead of just url? [14:40:01.0000] jyasskin: fwiw, filed an issue against Bikeshed here the other day https://github.com/tabatkins/bikeshed/issues/803 [14:40:19.0000] jyasskin: might want to reference in your PR so TabAtkins knows what's up. [14:41:04.0000] Still seems like argument names to Web IDL methods should never link to IDL terms defined elsewhere [14:41:49.0000] tobie: It's not really a bug in Bikeshed: https://drafts.css-houdini.org/worklets/#dom-function-arguments has `arguments`, which says any other use of Function/arguments should link to there. [14:42:38.0000] Houdini shouldn't be re-defining a name that's defined more authoritatively in WebIDL. [14:42:47.0000] jyasskin: I agree with Domenic, though. [14:43:22.0000] Is there never a case that you want to refer to a function's argument from another spec? [14:43:35.0000] I agree it's probably rare. [14:45:44.0000] I don't really have an opinion beyond wanting to get rid of a reference to worklets in WebIDL. :D [14:52:06.0000] I guess some specs do the thing where they have a table of arguments with a bunch of non-normative text about them :-/ [14:52:18.0000] Probably they should never export it [14:52:54.0000] tobie: I'm answering that issue right now. ^_^ [14:53:33.0000] Not a glitch - the Houdini spec defines a callback named Function already, so Bikeshed is assuming you want to link to it instead. [14:54:04.0000] I have code that ignores that for interfaces, and always defines (unless it's partial), but I'm generalizing it to all definition statements right now. [14:54:24.0000] TabAtkins: nice. [14:56:03.0000] And pushed. [14:56:57.0000] So now IDL blocks will always dfn; you only get a link if it's dfn'd elsewhere locally, or it's a `partial` construct [14:57:32.0000] Nice. [14:59:01.0000] TabAtkins: I must say filing issues against Bikeshed and seeing them fixed at such a pace is a pretty gratifying experience. [14:59:15.0000] Eventually you'll hit one that I let sit for a while. [14:59:28.0000] Thus why I have 70 open issues in the repo. ^_^ [14:59:43.0000] But I do like to get things fixed asap when I can. 2016-09-08 [22:46:19.0000] JakeA: I don’t attend the payment-apps meeting normally, so not sure if they were meeting this week or not [23:35:03.0000] tobie: oooh you did it [23:35:06.0000] tobie: congrats! [23:35:16.0000] tobie: and thanks for fast-forwarding the commits rather than merge commits [23:37:23.0000] annevk: yeah, didn't want to leave all other bikeshed users blocked until we shipped for too long. [23:38:16.0000] tobie: seems the last fixup commit isn't making use of bikeshed functionality () but I guess in general things are okay [23:39:24.0000] Oh, I have to look at that. Where is this? [23:41:43.0000] tobie: last commit: https://github.com/heycam/webidl/commits [23:43:19.0000] Oh. I hadn't seen it. [23:43:52.0000] I guess folks need to be schooled on the bikeshed way, now. [23:44:33.0000] tobie: seems https://github.com/heycam/webidl/issues/161 also wasn't closed as a result of fixing that [23:44:42.0000] Yeah [23:45:36.0000] Mmm. I'm much more worried about a chunk of an algorithm disappearing, tbh. [23:48:52.0000] tobie: well, yes [23:49:44.0000] I *had* to cut some corners, but this is really surprising. [23:54:26.0000] annevk: oh, it's this: https://github.com/heycam/webidl/blob/06bae58f/index.xml#L7667 [23:54:55.0000] I can see how that botched some of my regexes. [00:01:27.0000] OK, so I found three others places where the same thing happened. :( [00:02:02.0000] Will fix. Strongly hopping it's not the beginning of an avalanche or similar issues. :( [00:04:22.0000] tobie: I thought you used XSLT to generate bikeshed? [00:04:29.0000] tobie: XSLT has regular expressions? [00:06:19.0000] annevk: I did. [00:07:52.0000] annevk: but had to add a line by line post-processor to cleanup the indentation levels and markdownify the source, as XSLT's output can't really deal with this [00:08:30.0000] I'm not sure which of the two screwed those up. [01:04:55.0000] annevk: alright fixed those: https://github.com/heycam/webidl/pull/163 [01:05:20.0000] annevk: seems like a relatively well contained issue. [01:05:38.0000] Ah good [01:06:22.0000] annevk: combination of nested references and < > and related entities [01:07:28.0000] not sure where in the pipeline this broke and not sure it's really worth investigating either. [01:08:08.0000] heycam: what's your merging process for WebIDL? [01:08:47.0000] heycam: do you self-merge small fixes like these or get a second pair of eyes on them before you do? [01:14:36.0000] tobie: hi! I tend to commit my own changes without going through PRs unless I specifically want review from someone [01:15:46.0000] heycam: oh, nice. Less process. \o/ [01:16:00.0000] heh yes [01:16:17.0000] use your discretion I suppose [01:16:26.0000] I think bz errs on the side of getting reviews [01:23:18.0000] heycam: Well, as long as everyone feels confortable telling someone else they screwed up and need to revert, I'm all for favouring speed of protocol. [01:25:35.0000] speed OVER protocol. [01:27:08.0000] heycam: will you be at TPAC? [01:27:24.0000] tobie: no, won't be there [01:27:32.0000] :( [01:27:43.0000] :( [01:28:11.0000] heycam: not cool [01:29:32.0000] /me is not cool either [01:30:39.0000] Ms2ger: dude [01:31:13.0000] Well if it's going to be like this we might have to setup something else at some point [01:31:38.0000] yes [01:31:48.0000] we really should anyway [01:32:13.0000] Do one in Brussels and I'll consider it :) [01:33:31.0000] /me had forgotten that Ms2ger’s parole requirements prevent him for being able to travel outside Brussels [01:47:21.0000] MikeSmith: I'm going, btw. :) [01:56:54.0000] tobie: excellent [01:57:27.0000] tobie: so given that we should definitely schedule something concrete for WebIDL discussion [01:57:50.0000] Yeah, I pinged Yves. [01:57:54.0000] OK [01:58:17.0000] I have no idea how the super group is organized these days [01:58:24.0000] we can get a Tech Plenary breakout session of course, but that’d be just one hour [01:58:31.0000] oh [01:59:16.0000] yeah I forgot this was in the Web Platform WG [02:00:34.0000] Or, you know, folks could just go get coffee somewhere together. [02:00:38.0000] MikeSmith: I was thinking of: [02:00:39.0000] 1) short WPWG session just to update people on current status and plans [02:00:39.0000] 2) breakout session to gather requirements for future work [02:00:39.0000] 3) eventually ad-hoc meeting to get actual work done [02:01:09.0000] sounds good [02:01:16.0000] mkwst: yeah, see 3) above (so we get a room) [02:02:23.0000] mkwst: yeah agree and sometimes a lot more gets done in informal get-togethers between formal meetings [02:02:33.0000] "sometimes" [02:02:39.0000] heh [02:02:43.0000] I was being nice :) [02:03:09.0000] tobie: but what do you mean “we get a room”? [02:03:39.0000] MikeSmith: last time I looked, we could claim slots in rooms for ad-hoc meetings [02:03:49.0000] oh [02:03:56.0000] MikeSmith: think we did that for the sensor work last time [02:04:00.0000] OK [02:04:01.0000] booked half a day or so [02:05:30.0000] man the distinction between the bindings and the WebIDL stuff in there is just hard to grok in some places. [02:05:50.0000] that is why we should erase that distinction as much as we can [02:06:03.0000] How's an ES throw different from a WebIDL one? [02:06:21.0000] /me cries [02:07:11.0000] tobie, what do you mean? [02:08:14.0000] Ms2ger: http://heycam.github.io/webidl/#dfn-throw vs http://heycam.github.io/webidl/#ecmascript-throw [02:10:27.0000] I think it would be fine to redirect ecmascript-throw to dfn-throw [02:17:43.0000] tobie: the difference is that IDL still tries to be language-agnostic [02:18:05.0000] Kinda sorta [02:18:07.0000] "The meaning of each simple exception matches its corresponding Error object in the ECMAScript specification." [02:18:10.0000] tobie: I think we should give up on that to some extent [02:18:52.0000] We might actually short circuit some of IDL with wasm, but that should just become part of the IDL spec [02:19:07.0000] (E.g., a bunch of type coercion would no longer be needed.) [02:19:24.0000] Oh interesting [02:20:44.0000] Yeah, the extra level of indirection added by this near 1:1 mirroring makes things really complicated [02:22:28.0000] You have a {{SyntaxError}} IDL interface which has a SyntaxError name and maps the ES SyntaxError object. [02:22:46.0000] I'm not sure we're getting anything out of that beyond technical purity. [02:23:02.0000] I think it does make sense to have an IDL entry-point for all these things [02:23:34.0000] So if I define an IDL-algorithm, I use IDL-throw in it and IDL defines the implementation of that [02:24:18.0000] Basically IDL should define the language all spec-algorithms are written in [02:24:29.0000] How to do loops, comparisons, etc. [02:24:48.0000] A major outstanding issue is promise integration [02:25:00.0000] There's the promises guide, but that really needs to be part of IDL [02:25:35.0000] And once integrated IDL needs to really explain things in terms of ECMAScript, not hand-wavy language [02:27:22.0000] How much of that mean pulling in thing s described in HTML? [02:27:56.0000] (e.g. the infamous "in parallel" dfn everybody references) [02:30:39.0000] tobie: that might need to happen at some point [02:30:47.0000] tobie: baby steps [02:31:32.0000] tobie: but yeah, "in parallel" is a very special thing only available to IDL-algorithms and not really to ECMAScript [02:32:19.0000] annevk: oh yeah. Just trying to build a mental model of how the different pieces fit together so I can make educated decisions down the line. [02:33:11.0000] tobie: but IDL / HTML / DOM are intertwined [02:33:23.0000] tobie: Hixie would argue they should all be the same thing [02:34:02.0000] annevk: yeah, I'm just trying to figure out what the nodes are. [02:34:29.0000] tobie: points of overlap? [02:35:09.0000] yeah, I'm not exactly sure how to call these things yet. :) [02:35:39.0000] Let's not reuse node [02:35:52.0000] 😛 [02:36:15.0000] heh [02:36:41.0000] /me was thinking graph-theory node, fwiw. [02:37:26.0000] Fair [02:37:43.0000] tobie: next step is sorting through the bugs, right? [02:39:27.0000] annevk: if I do a window.open(url), does this create a fetch client for about:blank, then create another for the navigation to url? [02:39:41.0000] Or does the about:blank page use the client of the opener? [02:40:11.0000] annevk: There's a bunch of things [02:41:03.0000] annevk: removing the inline issues is one. [02:41:20.0000] JakeA: opener I think, but not defined yet [02:41:26.0000] annevk: triaging bugzilla is the second [02:41:52.0000] JakeA: and yeah, new one when you navigate [02:41:58.0000] annevk: ta. This matters for the reserved/target clientid stuff [02:42:05.0000] JakeA: