I made an accessibility error when coding up some tooltips. Here's where I went wrong, and how you can avoid making the same mistake:
https://jakearchibald.com/2026/my-tooltip-a11y-mistake/
Remote
Jake Archibald
@jaffathecake@mastodon.social
mastodon 4.8.0-nightly.2026-10-06Developer of sorts at Firefox, but this is my personal account. No thought goes unpublished. He/him.
0 Followers
0 Following
50 Posts
Joined November 01, 2016
GitHub:
Replying to
We now have an 'open web API' that requires you to comply with Google's "terms of use". I know many at Chrome disagree with this, but feel they can't say anything publicly out of fear of internal retribution. https://github.com/mozilla/standards-positions/issues/1213#issuecomment-4347988313
MacOS: When I restart, would you like me to re-open all these apps?
me: no
MacOS: hmm well that's a shame because I am going to do that.
me: ok
MacOS: Should I open them all maximised on the same desktop?
me: no
MacOS: hmm well…
Me: Argh I can't remember his name
@phae@status.fberriman.com: Who?
Me: Horrible podcaster
@phae@status.fberriman.com: Which one?
Me: he looks like he'd be called something like "Chud Gunt".
@phae@status.fberriman.com: Joe Rogan?
Me: yes thank you
MacOS: I see you've set Finder to appear on 'all desktops'
me: yes
MacOS: So when you cmd-tab to Finder, which desktop should I switch to?
me: well you shouldn't because…
MacOS: So I can switch to a random desktop? YAY!
Strangely satisfied to bag HTML issue 12345 https://github.com/whatwg/html/pull/12345
Replying to
@nachtfunke oh, Google had you all down as "strongly supportive" https://groups.google.com/a/chromium.org/g/blink-dev/c/iR6R7-nQeHI/m/gN4iEGEdAQAJ
Replying to
Because you can provide arbitrary image data to the encoder for the progressive passes, you're in full control over what those passes look like.
In this example, I've created a pass that's blurred except for the 'focal point' of the image (I'm not sure it's a good effect though).
Replying to
Of course, this means you can add shaders https://random-stuff.jakearchibald.com/apps/curved-markup/?vhs (again, Chrome Canary with canvas-draw-element flag)
Replying to
@slightlyoff@toot.cafe @jyasskin@hachyderm.io I think it's reasonable to complain about an abuse of position without offering an alternative.
Replying to
There are a number of optimisations the browser cannot do if you render html to a canvas, such as partial painting, optional painting, texture size management… so you should only do it if there's good reason.
Replying to
Live demo: https://random-stuff.jakearchibald.com/apps/partial-img-decode/?demo=cat-3-pass.avif&density=2 although the progressive rendering is Chrome-only right now.
Firefox bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1712813
The progressive passes work like frames of a video, in that subsequent passes can be inter-frames, reusing data.
Folks who are interested in popover=hint: have a play with https://random-stuff.jakearchibald.com/popover-hint/ and tell me anything that doesn't make sense. Or watch this:
I've proposed a series of changes to popover=hint. If you use web popovers, I'd love to know if the changes in the first post of this issue make sense, or, more importantly, if these changes would break your current usage of popovers.
https://github.com/whatwg/html/issues/12304#issue-4121948352
Replying to
@slightlyoff@toot.cafe I think you're underestimating the ToS issue. See the problem I laid out here https://github.com/mozilla/standards-positions/issues/1213#issuecomment-4347988313 If we allow APIs to have ToS, it creates an environment where developers are encouraged to block usage in other browsers in case they release with a ToS that'll get them sued.
Replying to
@slightlyoff@toot.cafe @jyasskin@hachyderm.io some thoughts on this on the other site https://bsky.app/profile/jakearchibald.com/post/3mkrt2wswpk2k
Replying to
@gundersen that would require a new feature to combine the resources, but it would be possible today with a service worker
Replying to
@slightlyoff@toot.cafe @ash@icosahedron.website …and that the API requires complying with a Google ToS. And that developer sentiment was (assuming good faith) poorly represented.
Replying to
@slightlyoff then how did this work? Did it ship by misrepresenting developers as "strongly positive"?
Replying to
@jyasskin@hachyderm.io @slightlyoff@toot.cafe yes, plenty of the negative reaction to the prompt API is from web developers.
Do you feel the evidence given in the I2S justifies saying developers are "strongly positive" about this feature?
Replying to
@joshhunt if there is then they failed to show it. They posted a couple of links, including a survey about something else.
Replying to
@Jespertheend well it's only a thing with this experimental feature. But yeah, you can upload any child of the as a texture (or draw it, in 2d canvas)
Replying to
@Jespertheend nice! But it sounds like you don't need pixel read-back in this case? The new API lets you use elements inside the canvas element as textures.
Replying to
@Jespertheend in Chrome, you can read the pixels back. It's how I make these videos https://mastodon.social/@firefoxwebdevs/116295985768415581 - it's mostly web content, captured with canvas, and piped to the video encoder API.
However, at Firefox, we're a little unsure about the readback stuff. What's your use-case for that in particular?
Replying to
In case you missed the intro to CloseWatcher: https://mastodon.social/@firefoxwebdevs/116290565797868795
Replying to
@vrugtehagel it uses a different model, but it supports progressive rendering https://jakearchibald.com/2025/present-and-future-of-progressive-image-rendering/#potential-progressive-rendering-with-jpeg-xl
However, Safari (the only browser that currently supports JXL) doesn't support progressively rendering JXL.
Replying to
@tylergaw@mastodon.social @jimniels@mastodon.social I think the problem with this is it looks like a browser setting, but it'll only do anything if the site supports it. DNT had similar issues.
Replying to
@duncan yep - this was my fault for making the picker display:grid, even when it wasn't 'open'. This gotcha has caught me out a couple of times. Fixed.
Replying to
@dpk I wasn't involved in WebUSB and can't remember the nature in which it shipped. But, at Firefox, we've been working on an implementation of WebSerial, and I'm absolutely supportive of that https://bugzilla.mozilla.org/show_bug.cgi?id=926940
Replying to
@jyasskin@hachyderm.io @slightlyoff@toot.cafe the I2S points to https://github.com/webmachinelearning/prompt-api/blob/main/README.md#stakeholder-feedback for evidence which includes "developer surveys", which is a link to a survey that appears to have asked if developers would be happy with this as an "extension API". So yeah, there's something a little fishy going on.
"Misrepresentation" is one of the things disallowed in Google's ToU for the Prompt API. I guess, therefore, summarising the I2S with the Prompt API would be breaking the rules.
Replying to
@trygve_lie @gundersen one thing that gets in the way a bit, is I think Chrome adds a timeout to avoid rendering progressive passes of AVIF in case the full file will be there soon.
Replying to
@vrugtehagel ohhh, which cross-browser differences have you seen? Although, differences in Safari don't count - they don't support popover=hint yet, so differences there are expected.
Replying to
Replying to
@jyasskin@hachyderm.io @slightlyoff@toot.cafe the reaction to https://github.com/mozilla/standards-positions/issues/1213#issuecomment-4347988313 and the responses on social media and Hacker News. What's the evidence for "strongly positive"? https://github.com/webmachinelearning/prompt-api/issues/74 was cited in the Intent to Ship.