In my experience LLMs are surprisingly bad at CSS beyond a very basic level. They work fine if you need to change the color of a button, but when it comes to actual styling work, even intermediate stuff like position:absolute or CSS grid, Copilot or even CC default to outputting correct-looking gibberish really quickly.
That's telling about CSS design. Folks here on HN are talking about how they purposely ask LLMs about APIs that don't exist, and they hallucinate with a better and more intuitive design that they would come up with on their own.
I don't know the best solution for the problem, but CSS is a very convoluted one.
Anchor positioning is part of Interop 2025. Firefox committed to shipping support for it this year: https://wpt.fyi/interop-2025
After that, it should take about 2.5 years for the feature to become Baseline widely available, and depending on your audience[0], you might be able to use it even sooner.
There are already a few sites that don't work properly in Firefox, people started testing only for chrome because its market share is so big.
Really unfortunate because it lets Google get away with anything they want, they are the new standard. But then again, I'm reminded of how Mozilla has pissed away all the users goodwill, and it's not a surprise.
I need a tooltip, with a pointer; but it seems that the current state of the spec does not allow for pointers; and most explainers studiously avoid this use case, as if this isn't a lion's share of what people do with anchored floating boxes.
Tooltips are normally visible on hover, so the pointer is your cursor. I've never added an additional arrow pointing to the element, nor had any designers ask me to do so. So I'd disagree that such a design is the "lion's share", but am curious what types of apps you create where you do find it to be so?
Someone in a comment below posted a link to Adobe Spectrum design system [0]. You will find similarly shaped tooltips in Shoelace [1], or shadcn [2]. The Popper library has it [3]. Github's design system has it (they call it popover) [4]. It's an extremely common design pattern.
They are using a stylized floating DIV (or something) not the built-in thing from the title attribute. Lots of design teams seem to want this, for consistency.
> Think a common approach is to just display a triangular svg beneath the tooltip
One killer feature of CSS anchor positioning is that it allows you to declaratively define fallback positions if the floating element does not fit into the preferred position. For example, you prefer your tooltips to appear below the anchor; but if the anchor happens to be at the bottom of the screen, there is no space below it, and so the floating element can flip to the top.
After the flip, the triangular svg will be pointing in the wrong direction.
I'm unsure what you mean by "pointer" - normally that just refers to the user's mouse cursor on-screen...
...do you mean you want a rich-HTML tooltip that is auto-positioned to ensure it's fully visible w.r.t. the browser's viewport but you also want the tooltip (or UI in general) to include an arrow shape that stays fixed on-target even if might be occluded by the browser?
Would be cooler if the whole system were more flexible: you simply define 2 anchor points (one on the target, another on the source, so center bottom would be bottom width 50% and top width 50%) instead of being limited to the 9 predefined areas
`position-anchor` is a high-level simple way of doing it, and it comes with the restrictions you mention. However, the `anchor()` function, which is also mentioned in the article, gives you the kind of flexibility you want.
I’m surprised it’s not in Firefox. I don’t remember the last time I ran into something in Safari and Chrome but not FF.
I was reading the article and thinking it would be a great thing to adopt for some code we recently wrote, but we have to support Firefox. And since we already have an existing solution that works, no point cleaning it up with this until Firefox adopts it.
I occasionally look into what CSS is being transcoded for the projects I work on, and it’s normally Firefox ESR that needs the most help. If you eliminate that from your browserlists configuration, your source and deployed CSS become a lot more closely aligned. For instance, it was only a year ago that Firefox ESR got CSS nesting.
It solves many of the pain points Tether[0] tried to solve.
For example it helps when the anchoring element is inside of an oveflow hidden/scroll container, but geometrically you need the tethered element to sit/extend outside of the container (so—for now at least—its DOM node needs to be outside of the container).
This always results in a ton of hacky JS to detect how the element should reposition itself if it overflows the screen (depending on the content and screen size)
This relies on being able to set the position relative to a parent selector, this doesn't work if the element you are positioning is not a descendant of the element you wish to anchor to.
Anchor positioning sounds cool, but I ran into some very unintuitive behavior when I tried to use it. Can’t remember the details, it was a couple years ago.
I guess you're being downvoted as a general nay-sayer, but you're right. I tried this feature last month and a bunch of browser bugs and design issues got in the way. I reported them, and they're being worked on https://github.com/w3c/csswg-drafts/issues/12466
The `margin:0` issue was particularly frustrating & imo should have been covered in the article, as it's a real gotcha when trying to use popover & anchor positioning in combination.
Yeah I could have mentioned the actual issues I had.
My first attempt was to anchor an element to another one that occurred later in the document order, and it didn’t work. The anchor must be placed before any of its dependents. It kind of makes sense, but doesn’t jump out as intuitive.
My problem is always been on sites that have a menu or something similar at the top. The anchor always inevitably goes to the very top of the screen gets covered by whatever menu it is.
At this point I'm just counting on LLMs to remember all the CSS specification cruft for me.
In my experience LLMs are surprisingly bad at CSS beyond a very basic level. They work fine if you need to change the color of a button, but when it comes to actual styling work, even intermediate stuff like position:absolute or CSS grid, Copilot or even CC default to outputting correct-looking gibberish really quickly.
That's telling about CSS design. Folks here on HN are talking about how they purposely ask LLMs about APIs that don't exist, and they hallucinate with a better and more intuitive design that they would come up with on their own.
I don't know the best solution for the problem, but CSS is a very convoluted one.
Unless there is a polyfill for Firefox, it will be at least a couple of years before you can rely on this for public sites.
Anchor positioning is part of Interop 2025. Firefox committed to shipping support for it this year: https://wpt.fyi/interop-2025
After that, it should take about 2.5 years for the feature to become Baseline widely available, and depending on your audience[0], you might be able to use it even sooner.
[0]: https://web.dev/blog/whats-my-baseline
There are already a few sites that don't work properly in Firefox, people started testing only for chrome because its market share is so big.
Really unfortunate because it lets Google get away with anything they want, they are the new standard. But then again, I'm reminded of how Mozilla has pissed away all the users goodwill, and it's not a surprise.
> Unless there is a polyfill for Firefox
Doesn't this count? Been there for several years.
https://github.com/oddbird/css-anchor-positioning
I need a tooltip, with a pointer; but it seems that the current state of the spec does not allow for pointers; and most explainers studiously avoid this use case, as if this isn't a lion's share of what people do with anchored floating boxes.
Tooltips are normally visible on hover, so the pointer is your cursor. I've never added an additional arrow pointing to the element, nor had any designers ask me to do so. So I'd disagree that such a design is the "lion's share", but am curious what types of apps you create where you do find it to be so?
Someone in a comment below posted a link to Adobe Spectrum design system [0]. You will find similarly shaped tooltips in Shoelace [1], or shadcn [2]. The Popper library has it [3]. Github's design system has it (they call it popover) [4]. It's an extremely common design pattern.
[0] - https://spectrum.adobe.com/page/tooltip/
[1] - https://shoelace.style/components/tooltip
[2] - https://ui.shadcn.com/docs/components/tooltip
[3] - https://popper.js.org/docs/v2/modifiers/arrow/
[4] - https://primer.style/product/components/popover/guidelines/
They are using a stylized floating DIV (or something) not the built-in thing from the title attribute. Lots of design teams seem to want this, for consistency.
Think a common approach is to just display a triangular svg beneath the tooltip:
https://react-spectrum.adobe.com/react-aria/Tooltip.html#exa...
> Think a common approach is to just display a triangular svg beneath the tooltip
One killer feature of CSS anchor positioning is that it allows you to declaratively define fallback positions if the floating element does not fit into the preferred position. For example, you prefer your tooltips to appear below the anchor; but if the anchor happens to be at the bottom of the screen, there is no space below it, and so the floating element can flip to the top.
After the flip, the triangular svg will be pointing in the wrong direction.
I'm unsure what you mean by "pointer" - normally that just refers to the user's mouse cursor on-screen...
...do you mean you want a rich-HTML tooltip that is auto-positioned to ensure it's fully visible w.r.t. the browser's viewport but you also want the tooltip (or UI in general) to include an arrow shape that stays fixed on-target even if might be occluded by the browser?
> I'm unsure what you mean by "pointer"
An arrowhead pointing at the anchor element.
Example: https://en.wikipedia.org/wiki/Tooltip#/media/File:Mobile_URL...
UPD: In spec speak, these are called tethers. The anchoring indicators
https://fantasai.inkedblade.net/style/specs/css-anchor-explo...
As mentioned at the end of TFA, Codepip's Anchoreum is an excellent way of learning this.
[0] https://anchoreum.com/
Would be cooler if the whole system were more flexible: you simply define 2 anchor points (one on the target, another on the source, so center bottom would be bottom width 50% and top width 50%) instead of being limited to the 9 predefined areas
`position-anchor` is a high-level simple way of doing it, and it comes with the restrictions you mention. However, the `anchor()` function, which is also mentioned in the article, gives you the kind of flexibility you want.
https://developer.mozilla.org/en-US/docs/Web/CSS/anchor
`position-area` syntax feels a little tough to remember, but I'm glad top/right/bottom/left is still available.
I was expecting boat anchors haha.
It being available on WebKit makes me hopeful for general adoption then.
I’m surprised it’s not in Firefox. I don’t remember the last time I ran into something in Safari and Chrome but not FF.
I was reading the article and thinking it would be a great thing to adopt for some code we recently wrote, but we have to support Firefox. And since we already have an existing solution that works, no point cleaning it up with this until Firefox adopts it.
Still, looks like a very nice feature.
> I don’t remember the last time I ran into something in Safari and Chrome but not FF.
It’s not especially uncommon. For instance payment requests, web share, and remote playback are all implemented by Blink and WebKit but not Gecko.
https://developer.mozilla.org/en-US/docs/Web/API/Payment_Req...
https://developer.mozilla.org/en-US/docs/Web/API/Web_Share_A...
https://developer.mozilla.org/en-US/docs/Web/API/RemotePlayb...
I occasionally look into what CSS is being transcoded for the projects I work on, and it’s normally Firefox ESR that needs the most help. If you eliminate that from your browserlists configuration, your source and deployed CSS become a lot more closely aligned. For instance, it was only a year ago that Firefox ESR got CSS nesting.
It's being actively worked on: https://bugzilla.mozilla.org/show_bug.cgi?id=1838746
> I don’t remember the last time I ran into something in Safari and Chrome but not FF.
IIRC Firefox lagged quite a lot on Color Profiles and :has
> I don’t remember the last time I ran into something in Safari and Chrome but not FF.
Background data sync/download with continuation
Any day now, Firefox.
> Any day now, Firefox.
Very true, they started 2 years ago and it has been constantly worked on with the latest update 12 days ago: https://bugzilla.mozilla.org/show_bug.cgi?id=1838746
So, it literally will be "any day now" :-/
Do we really need this? Why won't position: absolute and setting top/left/bottom/right suffice?
It solves many of the pain points Tether[0] tried to solve.
For example it helps when the anchoring element is inside of an oveflow hidden/scroll container, but geometrically you need the tethered element to sit/extend outside of the container (so—for now at least—its DOM node needs to be outside of the container).
[0] https://tetherjs.dev
Yes. Unless you want to rely on JavaScript libraries like popper and FloatingUI, we definitely need this for many use cases.
The simplest example is if you have content that it not contained by the box you're positioning against. Think tooltips, popovers, etc.
For some usecases like annotating content, this hugely simplifies things.
This always results in a ton of hacky JS to detect how the element should reposition itself if it overflows the screen (depending on the content and screen size)
That's fine for a lot of stuff. It becomes tricky to do certain other things. CSS-only tooltips are notoriously limited in scope.
This relies on being able to set the position relative to a parent selector, this doesn't work if the element you are positioning is not a descendant of the element you wish to anchor to.
Fundamentally no, html was fine. But hey it’s one fewer reason to reach for JavaScript, right?
Anchor post
Anchor positioning sounds cool, but I ran into some very unintuitive behavior when I tried to use it. Can’t remember the details, it was a couple years ago.
I guess you're being downvoted as a general nay-sayer, but you're right. I tried this feature last month and a bunch of browser bugs and design issues got in the way. I reported them, and they're being worked on https://github.com/w3c/csswg-drafts/issues/12466
The `margin:0` issue was particularly frustrating & imo should have been covered in the article, as it's a real gotcha when trying to use popover & anchor positioning in combination.
Yeah I could have mentioned the actual issues I had.
My first attempt was to anchor an element to another one that occurred later in the document order, and it didn’t work. The anchor must be placed before any of its dependents. It kind of makes sense, but doesn’t jump out as intuitive.
My problem is always been on sites that have a menu or something similar at the top. The anchor always inevitably goes to the very top of the screen gets covered by whatever menu it is.
Isnt there something like scroll-padding or scroll-margin? More or less an offset you can set so that doesn’t happen