WHATNOT meeting on 2025-09-04

[foolip]

What is the issue with the HTML Standard?

We had the APAC+EMEA WHATNOT call today. I agreed to chair it (first time) and unfortunately did not know to create an issue ahead of time. So here it is after the fact.

Next time, #11358 is a good template for an issue to be created well ahead of the meeting.

[foolip]

Agenda:

Review past action items

• None

Carryovers

• [zcorpan] Rendering <select> as a listbox is a one-line widget that opens a popup on iOS and Android #8189
• [foolip] Add the interestfor attribute #11006

New topics

• [zcorpan] Make <embed> and <object> behave like <iframe> for image loading #7939
• [dandclark] Proposal: Customized built-in elements via elementInternals.type #11061
• [keithamus] Proposal: Support image formats in <video> #11554

[foolip]

Minutes:

Rendering <select>

Simon: Don’t recall what needs doing. The request was for feedback, and that’s happened.

interestfor

Philip: Next up interestfor. Most checkboxes are checked- except implementer interest. CSS side has landed, :interest-source, :interest-target, and some properties to control delay. So when do we bounce things to stage 2? We do that by consensus of the companies of the steering group. How do we do this without Apple present?

Domenic: describes process document. We can consult apple async.

Philip: describes process document

Simon: We should try to get feedback from Apple in case they want to object.

Philip: For sure, but do we have Mozilla support for stage 2?

Simon: mozilla/standards-positions#1181

Keith: We have some concerns but no hard objections. Need to review the PR.

Philip: On device independence, we discussed this in OpenUI, provided a ton of options, spoke to UX folks. The model we landed on was an added option to the long press context menu. If you prevent the context menu from showing we immediately show the popover without additional delay. What’s covered in spec is keyboard, pointer devices (mostly mice & touch), and for any other modality it says the UA “should”. Not “must” because for example it might not make sense on a wearable.

Keith: How successful has the origin trial been?

Philip: I’ll get back with details.

Domenic: Looking at the WebKit standards position, if we’re happy with Mozilla actively affirming stage 2, then we should leave a note on the SP and ensure that if they strongly object then it is crystal clear.

Philip: Should we say stage 2?

Keith: I think we can provide positive affirmation of stage 2.

Philip: When should we make the comments?

Domenic: Tuesday

Philip: We can give it one week. I’ll also comment on the tracking issues, and on WebKit SP.

Embed and object

Simon: We made some changes, did some testing but it turns out there’s some file extension shenanigans. We need to assess the webcompat issues.

Domenic: Everyone wants to move to the autosizing model

Simon: SVG needs it, for webcompat.

Domenic: What if everyone just puts width/height on their embeds?

Simon: I need to look into it, I ran it on 1% or 1000 pages or so. Many object/embeds are flash. Running on full dataset is expensive but maybe we can bite that.

SELECT
 page,
 url,
 rank,
 REGEXP_EXTRACT_ALL(response_body, r"(?i)(<(?:object|embed)\s+(?:[^>]+)>)") AS elements
FROM
 `httparchive.crawl.requests` TABLESAMPLE SYSTEM (1 PERCENT)
WHERE
 date = "2025-08-01"
 AND client = "desktop"
 AND is_main_document IS TRUE
 AND REGEXP_CONTAINS(response_body, r"(?i)(<(?:object|embed)\s+(?:[^>]+)>)")
ORDER BY rank, page

Domenic: What we really need to find is objects & embeds with no width and height

Simon: I plan to filter it further.

Jake: Would it be width or height. Autosizing kicks in if it needs to calculate one or the other?

Domenic: Maybe. I don’t think Chome likes the extension behavior, I don’t think Mozilla does the same thing?

Simon: I think we figure it out after we’ve received something.

Domenic: You dont receive width or height until after it loads and then you have the content type.

Simon: Yes also the browsing context.

Domenic: So you always create the browsing context and add autosizing on response?

Simon: Except images dont have a browsing context. Maybe we create and destroy later.

Philip: Aren;t there magical documents that exist for this? Images and iframes. Is it same for embed and object?

Simon: Yes it is now in gecko but we have special code to hide the browsing context from the page.

Domenic: That still seems cleaner - hiding it from document.frames and such

Simon: If we can get rid of that too, that would be nice but we’re not sure on compat.

Philip: Can we interrogate this through http archive, or usecounter?

Simon: Usecounters seem nice, but I’m not sure exactly what to count.

Domenic: Upper bound could be page with iframe that access the index property, like frames[0] - you might get a different response.

Simon: It could break pages that fetch some other browsing contexts if you change the indexing.

Domenic: Adding a flag like “is hidden from index” isn’t the worst in the spec. We should spend some time on the minimum thing we can do, e.g. autosizing based on extension or content type. With spec editor hat on I’d be fine making everything hidden browsing context.

Simon: Not clear on handling of plugins, if a page tries to use flash, the page isn’t supported but it doesn’t have a browsing context.

Domenic: Enthusiastic about this though.

Philip: What’s different between iframes and embed/object with browsing contexts?

Jake: They’re the one thing on the platform where it may create a browsing context, based on the response, but in Chrome it’s based on extension?

Domenic: At spec level it’s much messier. Sometimes a browsing context, sometimes fallback content, sometimes image. A totally different path where it fetches the response and feeds it into the create a browsing context/fallback/image algorithms. Very different from anything else, an image for example goes through all the image request/response machinery, whereas embed does not.

Simon: Also serviceworkers don’t support object/embed because they’re too messy.

Jake: One reason behind that was browsing context, but the other reason is lack of trust of the security model of plugins. So even if we could know if there was a browsing context we’d still skip it because of plugins.

Domenic: There are no plugins on the web!

Jake: We can add a point saying if any UA is still doing plugins they MUST not or something.

Simon: Okay, I have some work to do.

Philip: Simon I shared a table with you - 118k rows. Is that right?

Simon: It’s not a lot of pages using embed/object.

Philip: Of 16 million in http archive?

Simon: 11 million for desktop.

Philip: Next up

ElementInternals type

Philip: Domenic wrote many words on the issue, can we do anything here?

Domenic: Does mozilla have words.

Jake: A frustration that custom element cannot be a form.

Domenic: This target submit buttons, generic buttons, labels.

Keith: Good motivations but lack of use cases.

Domenic: Normal buttons with popovertarget and so, but submit buttons, I’ve heard use cases…

Philip: As we don’t have much I’ll treat as new topic for next time. Any other late additions to the agenda?

Images in video

Keith: ex-colleagues at GitHub aren't sure image-in-video is useful for them. People don't want their markdown to be translated to <video>. They want a specific solution to if prefers-reduced-motion, gifs become click-to-play; maybe image-in-video is generalizing too far? Maybe UAs can just do that? Or we can put it in CSS? Any thoughts?

Domenic: Browsers don’t do stuff automatically to prefers-reduced-motion so they might not want to do it here.

Jake: Changing the default maybe not but an opt in.

Philip: Is it widely used enough?

Jake: prefers-reduced-motion comes from user but opt in needed by developer.

Simon: Some users have it without knowing right?

Jake: If someone is creating a css animation & video and we stop animating part of that it’s a tricky situation.

Domenic: Popular windows tip to improve performance!

Jake: CSS has animation-play-state, so making their own button if they want.

Philip: Theirs a whole slew of issues around this, autoplay is one, but seeking on gifs is another?

Jake: Lea’s proposal is making images work in a video tag, and an animated image format would get controls. I would like it if images/video were just raw media formats and an image tag could play any supported format.

Simon: Video formats in img, should we synchronise those animations. Gifs are synchronised.

Jake: Particular to gifs? Animated PNG? WebP?

Simon: All of them. I havent checked what Safari does though.

Jake: Animated AVIF in an img performs terrible compared to AV1 in a video. So something about that pipeline is difficult to optimise, I wonder if that’s part of that.

Philip: Image decoders I think are CPU only.

Jake: Gut feeling is videos in image tag are more interesting than images in video tag.

Simon: Videos are more efficient for network. Can transform to another format on the server.

Jake: CSS could be interesting, but how does SVG work? Maybe it doesn’t.

Philip: Do videos respect prefers-reduced-motion?

Jake: No, we discussed autoplay=maybe or something that would detect/respect it.

Philip: Is any user pref?

Jake: Yes but separate from prefers-reduced-motion.