Death to px, long live ch

8/24/2026 at 2:00:06 AM

The article links to

https://shkspr.mobi/blog/2018/11/the-myth-of-the-pixel-perfe...

which confuses me, because I definitely remember pixels being visible when I was growing up. I remember the first time I got a 1080p monitor. I felt a visceral terror at the fact that I could no longer see the pixels.

by andai

8/24/2026 at 3:48:28 AM

Yeah, you can see square-looking pixels on most screens if you display something with sharp aliasing. "There is no universal grid" is good to understand, but there are device-specific grids all over the place.

And all the examples that are ultra zoomed and/or showing antialiased font rendering obscure the pixels that are more visible in other situations.

That 0.01 to 0.04 degree range later in the article is big enough to drive a truck though. And you can see stairstepping at even finer distances.

by Dylan16807

8/24/2026 at 10:44:38 AM

It never says 0.04 degrees, just 0.04mm (corresponding to about 0.015 degrees at the stated 15cm distance).

by penteract

8/24/2026 at 3:34:42 PM

Ah, I see.

In that case let's look at what reaching .01 degrees means for a desktop. At normal viewing distance, a 20 inch screen needs to be 4k, and a 30 inch screen needs to be 6k. Very few desktops reach the level where pixels are mostly hidden.

by Dylan16807

8/24/2026 at 3:29:48 PM

Ultra zoom approximates a pixel grid, but the pixel grid it approximates no longer exists (and arguably never physically existed, it was always a mapping scheme). We tend to think of pixels as square bricks, but even in the CRT era they were closer to 4:3 rectangles than squares, and that too was an approximation because phosphor beams were a fairly analog technology and hitting a square (or rectangle) grid was an imperfect thing, especially with something as excitable as light. That's why popular arcade emulators have been picking up CRT filters that bleed neighboring pixels over each other to more accurately rebuild the CRT experience.

That's part of what led to dithering being as much or more important than exact color matching. That's also what led to advances at the time in sub-pixel anti-aliasing such as the earliest forms of Microsoft's ClearType brand which blended neighboring colors to make perceptually cleaner font rendering.

As we've moved to solid state hardware the grid didn't get more exact, and sub-pixels became even more interesting, rather than less. Most LED displays are triangular shapes, not square grids, with red/green/blue LED diodes packed closely together to map a virtual pixel. (To my understanding some LED displays get even more complex with hexagonal or octagonal packing.)

The device-specific grids even more rarely today reflect what we would expect of a "square" pixel grid. (Or even a 4:3 rectangular grid.)

Square cartesian grids are easy to program with it. It's always been a useful abstraction, even in the CRT era. It's easy to forget that it was always an abstraction. Ultra zoom operates on the abstraction, not the physical reality. It explodes a cartesian grid, because that's mathematically easy (and the useful abstraction we like to program with).

(All of that is before you get to things like OSes and Browsers gave up on mapping the virtual pixel cartesian grids to physical "pixel" grids in real life because the "pixels" became too small, so most users see "16px" fonts as 125-175% expansion to keep fonts readable without magnifying glasses. Web Browsers have "lied" about the physical size of pixels since roughly the end of the CRT era and browser pixels are font-relative too, in which case the other font-relative measurements are often more useful simply for being "honest". Nobody really wants "pixel perfect" accuracy in websites, not even the designers that pride themselves on it and think it matters more than other characteristics like responsive adaptability to device type.)

by WorldMaker

8/24/2026 at 3:54:28 PM

> Ultra zoom approximates a pixel grid, but the pixel grid it approximates no longer exists (and arguably never physically existed, it was always a mapping scheme).

Square pixels are the most real on LCD, and most desktops are still LCD.

Also when I said ultra zoom there I meant the actual photos showing enormous pixels.

> Nobody really wants "pixel perfect" accuracy in websites,

As long as you're keeping the scaling correct, snapping to the nearest pixel is sometimes useful. I don't know if it counts as a "website" but I've coded webgl shaders that adjust to the real device pixels and look much better for it.

125-175 percent expansion also happens to be the range where being pixel-aware can give you the most benefit. Fonts look good at those sizes because the browser can render at the final resolution instead of scaling afterwards. (Except when macOS renders at 2x and scales it down so everything is blurry)

by Dylan16807

8/24/2026 at 2:41:49 PM

Just to check: when were you growing up? In the CRT, it was often possible to see the physical pixels (i.e. the phosphors and the rows defined by the shadow mask), but there was not necessarily any guarantee that a logical pixel lined up with a physical pixel.

by amluto

8/24/2026 at 4:55:14 AM

> because I definitely remember pixels being visible when I was growing up

Was this on an LCD or CRT display? Because on a CRT, you can't really see the pixels [0] [1], so what you were seeing was probably the shadow mask instead [2]. You could definitely see the pixels on old LCD monitors though, and new LCD monitors too if you look really close, although even big LCD pixels aren't quite as prominent as the shadow mask is on a CRT.

[0]: https://en.wikipedia.org/wiki/Fixed-pixel_display

[1]: https://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf#page=8

[2]: https://en.wikipedia.org/wiki/Shadow_mask

by gucci-on-fleek

8/24/2026 at 9:45:34 AM

You can definitely see pixels on PC monitor CRTs where the shadow mask was much finer than the display resolution. PC monitors were not the same as TVs with nontrivial tech deviations between them.

by account42

8/23/2026 at 2:29:55 PM

I’ve found that ch and ex units are heavily influenced by latin characters. They just give weird results with non-latin characters leading to magic numbers. But the concept is really solid: use them if you want the spacing relative to text.

by microflash

8/23/2026 at 5:23:43 PM

This is a problem whenever someone outside a culture tries to design for that culture. I have just enough knowledge of say, CJK+Arabic to read them, but not enough to know if the font/kerning/justification I'm using is attractive or ugly to natives.

It's often easy to spot Chinese goods because English writing on the stickers or manuals is in that horribly ugly latin serif font with bad kerning, which almost certainly looks fine if you don't read English all day.

by qingcharles

8/23/2026 at 2:36:06 PM

Thought a magic number is when you hard code a number in logic instead of pulling from a data source

by LowTechHN

8/23/2026 at 2:43:47 PM

I'm guessing the GP means the weird results leads to adding magic numbers to get good results.

by goda90

8/23/2026 at 2:43:24 PM

Nobody uses ch, let's be honest.

by shevy-java

8/23/2026 at 3:48:06 PM

Until design tools like figma use these other units, people will be stuck to rems and pixels.

by dawnerd

8/24/2026 at 3:37:55 PM

rem is related to the ch, though. It's not an awful compromise if a designer insists on tools like Figma. Sometimes just taking their work in rems and just removing the `r`s is enough to meet their designs and also have better relative units.

(The difference between rem and em is "root", ie, only using the top-most font size and sometimes the font-size before things like zoom and other similar adjustments. If a designer complains about the `r`s you removed, sometimes just showing them the difference at multiple zoom levels is convincing enough.)

(The difference between em and ch is em is wider. If a designer is always giving you weird decimals like 0.79em or 1.24em that may be an indicator that they themselves would be happier working and thinking in ch. em is the width of the capital M which is often the widest letter in most latin sections of fonts. ch is the width of the number 0 which is often the most "typical" width in the latin sections of fonts.)

(ETA: Both of which relate to the comment above complaining that both ch/em and others fall apart when dealing with non-latin scripts, because they are established as latin-centric.)

by WorldMaker

8/24/2026 at 4:31:38 AM

Totally.

by JackCooper7

8/23/2026 at 2:56:06 PM

What I want to know is who decided to break px, the last thing we need is yet another screwball physical measurement, css was already full of them. But what "tut tut, everybody is using px and these new hi-dpi displays are rendering things really tiny, we can't have that" and now px as a useful unit is ruined.

Is there a way to use real pixels? My, admittedly quick, search says no. at least nothing jumps out at me from the spec.

by somat

8/23/2026 at 3:16:03 PM

px is still a useful unit. the user's operating system applying a scaling fator doesn't make it not useful.

if the user's operating system is configured to draw pixels at a 1:1 ratio, your pixels will be drawn at a 1:1 ratio. if the user's operating system is configured to draw pixels at a 2:1 or 3:1 ratio, you as a web developer don't get to override that, but you're still addressing pixels directly. if you want to write software that manages the user's hardware directly without giving them options to override it, the web is probably not the right platform for you.

by notatoad

8/23/2026 at 3:47:05 PM

The spec fafs around a bit talking about how the px is a "visual angle unit" But I am unable to form an interpretation where the "reference pixel" is anything other than 1/96 inch.

https://drafts.csswg.org/css-values/#absolute-lengths

I guess theoretically a tv could say "this tv will be viewed at 12 feet away so our reference pixel is 1/16 inch" and still be in spec. Which sounds like it has messy implications, Does anyone do this?

Anyway the point being a pixel should fundamentally be a hardware measurement, and if they wanted a "visual angle unit" they should have introduced one.(the moa?) But the last thing we need is a pixel being defined as a hard 1/96 of a inch. which is what we got.

by somat

8/23/2026 at 4:26:51 PM

Windows used to let you set the DPI setting of the monitor, circa Win95-WinXP. Software tended to only expect a value of 96dpi, so it wasn't a very practical setting to change, and eventually windows introduced display scaling (which leaves old software in the belief that you are on a 96dpi screen, while allowing software to opt in to knowing about the real device pixels)

Screenshot: https://www.softacom.com/wp-content/uploads/2022/03/11111-10...

I guess you could use that setting on an old version of windows to set a screen to 16dpi, and now a CSS reference pixel is 1/3rd of a real pixel? But more realistically, what the spec is getting at is that a mobile device should set its dpi setting close to the real device setting, and a reference pixel is now something that looks about as big as a pixel on a 96dpi screen. That nobody actually sets their screen dpi on desktops is probably of secondary importance to them

by wongarsu

8/23/2026 at 3:51:25 PM

what the spec says doesn't really matter though, the spec is just nonsense people wrote down to try to back-explain decisions that the browsers had already made.

pixels are pixels, and no browser implements them as a fractional-inch measurement. in every actual implementation i'm aware of, they're a hardware based unit.

by notatoad

8/23/2026 at 3:58:50 PM

They aren't hardware based. As I say in the first link, pixels mostly don't exist in hardware.

Screens very rarely contain a matrix of individual squares.

For example - https://global.samsungdisplay.com/29043/

by edent

8/24/2026 at 7:19:22 AM

You’ve linked one example of exactly how those samsung displays do consist of a matrix of individual squares.

Here, samsung draws a border around one: https://global.samsungdisplay.com/wp-content/uploads/2021/11...

A pixel is made up of multiple components because everything is made of multiple components.

by systoll

8/24/2026 at 9:18:19 AM

Look at how they describe it on that page:

> Samsung Display succeeded in designing the most optimal spatial frequency design with a 45 degree diagonal symmetry across the entire display layout.

See the image: https://global.samsungdisplay.com/wp-content/uploads/2021/11...

All of the components are in that diamond pattern. When you try to draw to a specific pixel width, the display chooses which sub-pixels make up a logical pixel. There is no specific square for you to target.

by edent

8/23/2026 at 6:10:15 PM

Pixels very much exist in hardware outside mobile. Microsoft’s clear type even exploited the specific layout of color channels in discrete hardware pixels. Mobile displays are all over the place, but most other displays use standard RGB layouts.

by jonhohle

8/23/2026 at 7:24:02 PM

You've described the exact problem. Sub pixel rendering exploits the fact that most display pixels aren't discrete logical squares.

https://en.wikipedia.org/wiki/Subpixel_rendering

by edent

8/23/2026 at 7:42:39 PM

Right, but there's still an addressable set of display units that convert an RGB signal to actual glowing lights, right?

It seems like defining px to be perceptual is one of those compatibility hacks that probably helped a ton back in the day, but we have perceptual units and now we're stuck with that baggage.

I don't have much of an opinion on this (I use em) but it has taken me by surprise a few times.

by SR2Z

8/24/2026 at 9:56:38 AM

> Right, but there's still an addressable set of display units that convert an RGB signal to actual glowing lights, right?

Yes. But those glowing lights aren't little squares in a neat horizontal and vertical rows.

The square pixel is useful in logical contexts - like an SVG or bitmap graphic. But when it is turned into a physical display, those pixels aren't displayed as little squares. The RGB elements are arranged in dozens of different formations depending on the display technology.

by edent

8/23/2026 at 4:43:00 PM

You can kinda get close in pure CSS with CSS media queries using the resolution query on dppx measurements; but that only allows you to be as granular as however many distinct "dots per CSS pixel" values you want to add media queries for. Something like this would let you scale by device resolution for both traditional 96dpi screens, higher DPI 192dpi screens, and super high resolution 288dpi screens:

    body { --resolution-scale: 1; }
    @media (min-resolution: 2dppx) { body { --resolution-scale: 2; } }
    @media (min-resolution: 3dppx) { body { --resolution-scale: 3; } }
    .myPixelSizedElement { width: calc(96px * var(--resolution-scale)); }

But unfortunately there's no way to directly use the devicePixelRatio value you can read from Javascript directly in a CSS calculation to be able to get a precise scaling on any arbitrary pixel density.

by drysart

8/23/2026 at 5:08:12 PM

> But unfortunately there's no way to directly use the devicePixelRatio value you can read from Javascript directly in a CSS calculation to be able to get a precise scaling on any arbitrary pixel density.

The indirect way is for the JS to post the metric to the server, then you run with server-generated CSS. Don’t ask me if it actually works well though.

by cobbzilla

8/23/2026 at 5:25:55 PM

Much simpler, you also just have JS update (and keep updated - initial DPI is not always the current DPI) --devicePixelRatio under :root on the client side. Something like that existing by default is all that's really missing.

by zamadatix

8/23/2026 at 3:22:31 PM

You could set the CSS zoom property to 1/window.devicePixelRatio with JavaScript but I think you can't do it with CSS only. If you do that then px is physical px. This doesn't include browser zoom though or mobile pinch to zoom. You could also get those with JavaScript but then people would be really confused by the behavior of your website.

You definitely can't get pure integers though it's always a float even if you use whole numbers. It does eventually get rounded before becoming pixels though.

by fsmv

8/23/2026 at 4:07:39 PM

That is sort of where I ended up, Not that it matters, Pixels should probably never actually be used. It just offends me that a unit that is sort of messy and tied to physical hardware but does provide measurement that can not be found elsewhere. gets neutered and we end up with another useless absolute unit, inch, point, millimeter, pica, the Q(did we really need the Q?) and now the pixel.

by somat

8/23/2026 at 5:00:09 PM

To be honest, I'm quite thankful to whomever decided this. If it were the other way around and everyone used px but it represented true pixels, there's no doubt developers would be incorrectly designing interfaces across the board and websites looking totally off would be a common occurance due to you having an hdpi display or not having one (like what frequently happens with x based apps in Linux).

As an aside, technically, yes, it may have been better to converge on some more clearly "fake" unit rather than calling it a pixel but for me that's a nit pick.

by _matthew_

8/24/2026 at 9:50:23 AM

Unfortunately people are designing interfaces incorrectly anyway. Just look at the font size differences between websites from the 90s and today.

by account42

8/24/2026 at 3:20:18 AM

Who broke px? It was W3C. In "CSS Values and Units Module Level 3", px was defined as 1/96th of an inch

But even in CSS1 (1996), a "reference pixel" based on a typical 90 DPI monitor was recommended. You could argue its always been broken

by culi

8/23/2026 at 3:24:37 PM

No one decided to break it, it was 1/96th of an inch from the beginning.

The problem with different screens having different pixel densities was already present and obvious in 1996.

"Real" pixels are useless if you don't know anything about the screen you're going to render on.

by cedilla

8/23/2026 at 3:39:52 PM

It was around 1/96" on a conventional PC screen because that was the conventional pixel density for PC screens; it was 1/72" (i.e. 1pt) on a conventional Mac screen for the same reason. That was something to deal with already in the GIFs-in-tables era and remained one as we moved on to CSS, right until Retina happened.

Unfortunately, Retina gave us a nonsensical and hacky “scaling ratio” rather than separate and independent units of physical length, viewing angle, and interaction granularity. (I still can’t get an A4 PDF in GNOME to display at “100%” such that it would actually be 210mm wide on my screen. And Darktable is the only piece of software I know where you can configure ppi and ppd separately.)

by mananaysiempre

8/24/2026 at 2:37:45 AM

Knowing about the screen also doesn't do you much good if you don't know how far it is going to be away from the viewer. (Yay at fonts rendering with 3 point on my projector because it reports some 6m³ of "screen" area.)

by krautsauer

8/24/2026 at 9:52:53 AM

For most use cases yes, but there are some where you could want something on screen to have the same size as a real physical object.

by account42

8/23/2026 at 3:28:25 PM

Correction: in CSS1, from December 1996, a pixel was defined slightly differently, which works out to 1/90th of an inch.

by cedilla

8/23/2026 at 5:39:14 PM

They decided to break px because web designers used it for things like setting text size. Why did designers use px rather than pt which was supposed to scale? Because early browsers interpreted pt differently across platforms, so specifying text size in px rendered more consistently.

by bazoom42

8/23/2026 at 2:46:06 PM

I also did some experiment with ch many years ago. I found that 60ch is ideal width for block text for easy reading. too bad it is pretty hard to make websites with only 60ch wide.

by npn

8/24/2026 at 2:03:38 AM

Why's it hard? I just tested it by applying 60ch width to TFA. It hardly changed, it was already not much wider than that.

This website layout (tall brick surrounded by voids) was very popular in the 2000s.

by andai

8/24/2026 at 7:04:24 AM

Huh?

  <main style="max-width: 60ch; margin: auto>
    Your content
  </main>

What is hard about that?

by atoav

8/24/2026 at 4:30:13 PM

what hard is making a fully featured website, with panels no wider than 60ch.

typically 60ch equal to 480px (font size 16px), so you need sidebars to the left and the right. which is fine, the holy grail was like that. but if you want to design a layout that look good in all screen resolutions, then a fixed layout does not work. you would have a dozen compromises and layout switches to make it work. at that point, forcing a block to have at max 60ch is painful.

and, if you do the main + sidebar layout, with maximum width for the whole site is around 960px (or 1200px if you have multiple sidebars) then suddenly you do not need to think about 60ch width anymore, because it naturally just fits (except for some width points)

by npn

8/24/2026 at 7:41:44 PM

main should contain the semantic content of the website, you can still make full width headers and footers and have on the side. Well and <aside> can be just where the name implies, depending on the screen width.

This is really not that hard nowadays. How to do responsive design is no black magic and can be done with only HTML and CSS.

You basically just need to wrap (no pun intended) your head around a few basics. Especially CSS grid templates combined with media queries allow you to rearrange the whole page depending on the screen width in a simple way, especially compared to how we had to do this 20 years ago.

You write your website semantically (or ensure the template produces semantic HTML) and then you write the CSS with realtime preview right there in the browser.

by atoav

8/25/2026 at 12:30:36 AM

? Why are you assuming that I don't know about that.

Actually have you really design practical webapps before? Because your post only have empty pretty words with no substance.

Semantic html tags are just semantic, it is irrelevant for the discussion.

Responsive design is hard, have you ever try to reize your windows and see how your pretty website behaves? Leave aside stupid people thinking you only need to care about phone device (in portrait mode even) and laptop, most people also think 3 or 4 device widths is enough. Sure some smart one already promoted to break the screen by content, not fixed device size, but even that comes with some downsides.

To make a website works decent on all the possible screen sizes is a lot of work, I spent like 4 months updating and fixing my webapp when I first developed it.

by npn

8/24/2026 at 9:56:18 AM

Just leave the viewport size up to the user rather than making assumptions about which widths work best for people you don't know.

Also despite the aggressive title, this article gives basically no reason for why you shouldn't use px.

by account42

8/23/2026 at 4:12:14 PM

I'm still waiting for operating systems to adopt true physical units. You know, that trick PostScript printers have doing since the 80s (running at 300dpi no less). I realize there are large screen projectors in the wild, that's solvable problem. Bitmaps icons are also an issue, but again, solvable.

by jeberle

8/24/2026 at 2:03:25 AM

Real physical units make sense when you're printing on paper.

I don't see how they make sense on screens. A one square inch image is fairly large on my phone, not too big on my desktop monitor, and pretty darn small on my tv. That's not to say that pixels make sense to use for layout at those scales either, but before everybody got into high dpi with scaling, it kind of worked... If you were comfortable with tiny text, you ran a higher resolution and if you needed things big, you ran a lower resolution.

by toast0

8/24/2026 at 10:01:28 AM

> That's not to say that pixels make sense to use for layout at those scales either, but before everybody got into high dpi with scaling, it kind of worked... If you were comfortable with tiny text, you ran a higher resolution and if you needed things big, you ran a lower resolution.

And even with so-called "dpi" scaling, CSS pixels remain that except users have more options to also have their large TV display finer details.

by account42

8/23/2026 at 4:33:54 PM

I’d like to hear the author’s approach to vw and vh units. I’ve tended to use these units where I want a specific feel and it’s worked relatively well over the years. I’m surprised it was not mentioned given the justifying of ch was around the desire of a consistent layout.

by squidlib

8/24/2026 at 3:50:19 PM

Consistency in viewport size versus consistency in relation to the width of text. It's a little bit of a "top-down" versus "bottom-up" distinction in thinking. (I often find the best answer is "both". vw/vh for "big picture" alignments like outermost grid structures and ch/em/lh/ex for "content" work inside those big grids.)

by WorldMaker

8/23/2026 at 2:55:58 PM

What happened to em?

by cynicalsecurity

8/23/2026 at 3:24:47 PM

That's what I wondered. I found the spec*. I'd vaguely thought that 1em was the width of "M" but apparently it's just the font-size, and the connection between font-size and the actual size of the text is very font dependent.

So 1ch is defined to be the advance width of "0" (zero), so it's actually telling you something about the current font.

* https://www.w3.org/TR/css-values-4/#font-relative-lengths

by masfuerte

8/24/2026 at 3:45:38 PM

Even when em is literally the width of "M" and ch the width of "0", the difference is often pretty stark, with "M" being most commonly the widest character in the latin set and "0" being among the most "typical"/"average" in width.

by WorldMaker

8/23/2026 at 6:27:19 PM

I know this is an irrelevant factoid and literally does not matter at all, but it's a tad more meta that just em = font size. Prepare to be bored:

The em unit/point size in typography was based on the height of the lead sorts you were using. [0] (A sort is basically a little lead stamp with a single letter jutting out on the front face. [1]) Its almost always larger that any dimenion of the actual character, since the sorts are just clamped together when doing the actual typesetting, so you want a gapless grid of lead to keep everything in place. [2] You couldn't have any spacing between letters/lines if the castings themselves were only as tall or wide as the character.

That meant the em height was a real physicial thing you could measure with a ruler, commonly in points, where 1/72 of an inch = 1 point. When you ordered typefaces from foundaries, you'd specify what point size you wanted, so all of your large-point headings would line up in one slug, which you'd clamp above another slug of body copy text all sized to a smaller point size. That tied an em to both the size of text, but only in the context of 1 font family. You could leave the em size the same and cast a glyph larger on the same size of sort, changing the baseline, line-height/ledding and letter-spacing/tracking. Now it doesn't line up with your other font faces with the same em size, even if they clamp together just fine.

But in modern typesetting, you don't have a physicial em-height lead block to deal with. Internally, the font uses it's own unitless values to describe both the vector shape of the glyphs, as well as units like the line height, letter spacing, and em height. It's no longer a "real" measurement you could actually measure. It's a relative one that acts like the unit value when scaling fonts. [3] If you render a font at 12pt, the text rendering scales the relative em unit until it would be 12pt or 1/6in, and then apply that scaling factor to all glyphs.

That means, digital fonts are emulating the same weird use for ems as a proxy for both the size of the glyphs AND the size of the lines/spaces between characters. If you shrink the internal em unit inside a font file, the size of glyphs increase, but so will the line height/letter spacing and other font metrics.

So... all of that together, yep the em unit is 100% dependant on the font's own settings, but it's not the font size itself. It's a sort of base-unit for everything else in the font, where 1em is defined as being a certain height in the internal font domain AND being the height of whatever the font renderer says the point size is. If you halve the em height and double the point size, you end up with exactly the same raster ouput. It also means you can't look at the screen and find any part of the letters/lines that is precisely 1em, like you could for simplier glyph-size-based unit like 1ex or 1ch. (Unless there's an em-dash... but even then, making the em-dash exactly 1em long is just convention.)

[0] https://www.thomasphinney.com/2011/03/point-size/#:~:text=ou...

[1] https://en.wikipedia.org/wiki/Sort_(typesetting)

[2] https://commons.wikimedia.org/wiki/File:Sorts_on_composing_s...

[3] https://www.thomasphinney.com/2011/03/point-size/#:~:text=Ty...

by graypegg

8/24/2026 at 2:38:50 AM

em is a generic "font-size" without regard to the axis. ch is specifically related to the font width, and ex is specifically related to the font height. So if you want things like padding and margins to be proportionate to the font size, then using em will be suboptimal for fonts that happen to be tall or short. And as a web developer, you generally don’t know which font will be used (content management systems, designers making different choices in the future, users overriding fonts for accessibility reasons, etc.)

For lengths that you specifically want to be axis-independent and related to the font size, e.g. border-radius, then you should still use ems.

by JimDabell

8/23/2026 at 11:44:38 PM

I always use em or px, basically nothing else (except perhaps vw or vh). When something should be relative to the size of text, it's em. When it should be relative to the size of images, it's px.

by BSTRhino

8/23/2026 at 5:10:37 PM

designers don't like doing math

by micromacrofoot

8/24/2026 at 2:36:20 PM

It's starting to look like win32 was the pinnacle of GUI API design.

by BearOso

8/23/2026 at 3:37:43 PM

content-driven websites that aim to be responsive and accessible should use as little size-related styling as possible. fr for containers and rem/em for text.

by theokrueger

8/23/2026 at 5:41:18 PM

I’m somewhat out of the loop but I remember the time everybody switched from px to em’s. When I came back to CSS for a short while it was px again and behold it worked for 4K screens. To be honest I didn’t really liked the usage of em and actually prefer px.

by larusso

8/23/2026 at 2:43:04 PM

Layout shift when the font loads

by rado

8/23/2026 at 3:03:53 PM

Not necessarily. See https://screenspan.net/fallback

by microflash

8/23/2026 at 4:14:25 PM

Headline length changes in that demo: https://screenspan.net/prototypes/fallback-font/

by rado

8/24/2026 at 4:35:25 AM

why would anyone care if 1cm is actually 1.024cms on some monitors?

by coolThingsFirst

8/24/2026 at 11:01:39 AM

I tried and with 1:1 zoom, a "10cm" width DIV was actually less than 7cm. Why even call it "cm" if it is actually "whatever, I don't know"...

by moring

8/24/2026 at 9:50:17 AM

Maybe a website is selling screws and wants to show a 1:1 picture.

by Zecc

8/23/2026 at 3:19:48 PM

I will not take CSS advice from a website that looks like such utter crap on mobile

by awestroke

8/23/2026 at 3:52:08 PM

May I ask which browser you're using? I do test on all the major ones at a variety of screen sizes.

Or, have your perhaps selected one of the different themes at the top of the site?

by edent

8/24/2026 at 10:28:52 AM

Agreed. This site is always literally unreadable for me on mobile like they placed a blur filter over then entire page.

by SoMomentary

8/23/2026 at 9:09:46 PM

[dead]

by blueshoess

8/23/2026 at 3:49:58 PM

[dead]

by LowTechHN

8/23/2026 at 2:42:54 PM

There is no way I will abandon px.

In fact, I think CSS made the wrong decision by proliferating so many things, em, %, px, ch, whatever-else. The human brain is not well-equipped to have so many things for basically the same thing. That's a design flaw in CSS, plain and simple. And the author is thus also wrong. There won't be a mass movement of people using "ch". It would be a good first april article though.

by shevy-java

8/23/2026 at 4:31:37 PM

I’d like to hear the

by squidlib