| 06:19 | <MikeSmith> | annevk: I’m working on implementing a document-headings builder based on the spec text in the PR #3499 patch, and one difference I’ve noticed is that unlike the old outline algorithm, it doesn’t seem to take into account whether any ancestors of a heading have a `hidden` attribute |
| 06:19 | <MikeSmith> | (nor whether the heading itself does) |
| 07:35 | <annevk> | MikeSmith: should that affect whether it’s a heading, basically? |
| 07:35 | <annevk> | MikeSmith: that would add some complexity |
| 07:36 | <MikeSmith> | yeah I know ー it’s part of what made the old outline algorithm more complex |
| 07:37 | <MikeSmith> | but to answer your question, of course it doesn’t affect whether it’s intrinsically a heading |
| 07:37 | <MikeSmith> | but in practice, I’m not sure that’s what’s important |
| 07:37 | <annevk> | MikeSmith: so what does it affect? |
| 07:38 | <MikeSmith> | well, does a screen reader announce a heading if it’s hidden? |
| 07:39 | <MikeSmith> | I thought a big part of the goal of this was to expose heading levels to AT in the right way |
| 07:39 | <annevk> | MikeSmith: in that case it affects whether it’s a heading, right? Though AT also respects display:none so hmm |
| 07:40 | <annevk> | Anyway, we could make this work and impl presumably do it already |
| 07:41 | <MikeSmith> | for h2-h6 yeah I guess implementations do |
| 07:41 | <a-ja> | MikeSmith, offscreen headings do get announced unless aria-hidden or role=presentation, generally, iirc |
| 07:41 | a-ja | is no expert |
| 07:41 | <MikeSmith> | yeah? |
| 07:42 | <a-ja> | ...but i've studied some |
| 07:44 | <MikeSmith> | OK well anyway right now I’m just trying to consider it in terms of what I need to do in order to replace the backend for the current Show Outline feature in the HTML checker with one based on the PR #3499 patch |
| 07:46 | <a-ja> | url? |
| 07:46 | <MikeSmith> | and with the current backend of this, which is based on the old outline algorithm, headings in subtrees with `hidden` attributes don’t get included/shown |
| 07:46 | <MikeSmith> | a-ja: URL for what? |
| 07:46 | <a-ja> | the PR |
| 07:46 | <MikeSmith> | ah |
| 07:46 | <MikeSmith> | https://github.com/whatwg/html/pull/3499 |
| 07:46 | <a-ja> | tks |
| 07:48 | <MikeSmith> | annevk: thinking about it more, maybe the broader issue/difference is the patch doesn’t define anything about how to build an outline |
| 07:48 | <MikeSmith> | or a set of heading levels or whatever we would want to call it |
| 07:49 | <MikeSmith> | it defines the term "document headings" but doesn’t say anything about doing something with it |
| 07:50 | <MikeSmith> | doesn’t say anything about exposing "document headings" as such, as a set |
| 07:50 | <MikeSmith> | or rather, as a list I guess |
| 07:51 | <a-ja> | i dunno how screen reader users prefer to nav around a site these dats...whether by headings or landmarks (which may have aria-label) |
| 07:52 | <a-ja> | *days |
| 07:52 | <MikeSmith> | a-ja: both, I htink |
| 07:52 | <MikeSmith> | since it’s not an either-or |
| 07:53 | <MikeSmith> | I guess it probably depends a lot on the site and how it’s marked up |
| 07:54 | <MikeSmith> | but one thing I do know is that in the years of discussion we’ve had about this, there have been lots of comments from SR users saying they do often navigate by headings |
| 07:54 | <MikeSmith> | not saying that they do exclusively, but saying that it’s comment |
| 07:54 | <MikeSmith> | *common |
| 07:55 | <a-ja> | lately i've been going by whether it's a heading i'd put in a toc...real content, whereas nav's get aria-labels instead of heading |
| 07:56 | <MikeSmith> | annevk: so this point, as far as I what I need for my purposes, I guess I just have to assume that `hidden` headings should not be including in the “document headings” view/outline that the checker exposes |
| 07:56 | <a-ja> | it keeps the a11y checker tools happy, but i've really no idea of any surveys which ways screen-reader users prefer |
| 07:58 | <MikeSmith> | annevk: I mean, since the the spec doesn’t explicitly say what should happen for `hidden` headings ー nor no longer say anything explicitly about how to build an outline |
| 07:59 | <MikeSmith> | a-ja: for us I’m not sure it’s so relevant which they prefer, since we need to support both |
| 08:00 | <MikeSmith> | I mean as far as having the spec define how UAs expose heading levels |
| 08:01 | <a-ja> | gonna create an <h> element, too, while you're at it? :) |
| 08:01 | <MikeSmith> | nope |
| 08:01 | <a-ja> | heh |
| 08:01 | <MikeSmith> | definitely not going to do that |
| 08:02 | <MikeSmith> | in hindsight, we never should have created <section> or <article> |
| 08:02 | <MikeSmith> | especially <section> |
| 08:04 | <a-ja> | not so sure about that...certainly nudged a lot of folks (myself included) into more semantic structure |
| 10:57 | <annevk> | MikeSmith: we ahould figure out what to do with hidden I think |
| 10:57 | <annevk> | MikeSmith: we did for main |
| 10:57 | <annevk> | MikeSmith: I guess we can be more explicit on doc headings too somehow |
| 11:30 | <MikeSmith> | annevk: dunno maybe it’s not actually important ー I mostly just noticed it as a difference compared to the previous requirements |
| 11:31 | <MikeSmith> | annevk: but anyway, speaking of differences, if I read the patch text correctly, <hgroup> never takes it heading level from the level-ness of its child h1-h6 but instead takes it only from its nesting depth? |
| 11:32 | <MikeSmith> | if so, that strikes me a very suboptimal |
| 11:34 | <MikeSmith> | I think <hgroup> was never really anything more than a hack to suppress sub-section creation in the old outline algorithm |
| 11:34 | <MikeSmith> | so given that we’re discaring the old outline algorithm, I think we should consider just essentially ignoring <hgroup> altogether |
| 11:36 | <MikeSmith> | what I mean is, the spec should say that <hgroup> should not be treated as a heading, but instead its child heading elements should take whatever level they otherwise would if they had not <hgroup> parent |
| 11:37 | <MikeSmith> | that’s just as easy for UAs to implement (doesn’t require keeping any more state than what the current patch says, doesn’t require looking down/ahead) |
| 11:39 | <MikeSmith> | in other words, if we’re dispensing with the idea that headings create sections (as they did in the old outline algorithm), then we no longer have any useful function/behavior to associate with <hgroup> |
| 11:42 | <MikeSmith> | it can instead become purely just some extra markup wrapper that’s effect-less/harmless but just there for legacy reasons |
| 11:43 | <MikeSmith> | or alternatively, we can deprecate or obsolete it |
| 11:47 | <MikeSmith> | the use-counter stats from the HTML checker showed <hgroup> being used in only ~0.2% of documents, if I recall correctly |
| 12:14 | <annevk> | MikeSmith: how would we do subheadings then? |
| 12:15 | <annevk> | MikeSmith: the current idea is that hgroup is the heading in such cases and UAs can represent the multiple headings nested therein as of being increasingly less important |
| 14:49 | <benjamingr_> | Hmm, would be really useful to have a highWaterMark in `fetch` |
| 15:36 | <Domenic> | benjamingr_: file a bug? Seems pretty easy to add in theory. |
| 15:54 | <benjamingr_> | Domenic: cheers https://github.com/whatwg/fetch/issues/689 |
| 18:58 | <Zegnat> | Quick question. In 14.3.1 (https://html.spec.whatwg.org/#hidden-elements) what does the `i` in `input[type=hidden i]` mean? |
| 18:58 | <Domenic> | Zegnat: case-insensitive |
| 18:58 | <Zegnat> | Quick question gets a quick question. thanks Domenic! |
| 18:58 | <Zegnat> | quick answer ... ugh |
| 19:44 | <espadrine> | Hi, is there a way to ask the browser not to share same-origin XHR with withCredentials set to true? |
| 19:45 | <espadrine> | Access-Control-Allow-Credentials:false does not do it. |
| 20:21 | <domfarolino> | espadrine: What do you mean by "not to share" |
| 20:26 | <espadrine> | domfarolino: as in, either not send the credentials with the request, or not yield the response body to the xhr / fetch |
| 20:37 | <domfarolino> | espadrine: Only on same-origin requests? |
| 20:39 | <domfarolino> | with fetch() you can set a Request's mode to `no-cors` (and credentials = `same-origin`/`include`) which will give an opaque response if that makes sense. |
| 20:51 | <espadrine> | domfarolino: In a situation where I don't trust nor control what a specific page contains, and that page comes from my origin. I can't expect the page to simply not do the harmful thing and set credentials as I want. I don't want it to be able to make a request to my server with the credentials (which it can't read since they are secure cookies). |
| 20:57 | <Zegnat> | Hmm, I tried to submit the Participant Agreement (for PR whatwg/html#3606) but it gave me an Error 400 “Invalid 'scope-workstreams' field”. Known error? |