If you could start over from scratch, how would CSS work?

ben profile image Ben Halpern ・1 min read

CSS has a lot of issues. Now that we have a few decades of knowledge, how would you do things differently.

And I'm talking about web styling in general, doesn't have to be "cascading stylesheets" or anything like what we currently have.


markdown guide

Wait, people don't like CSS? To me, CSS actually seems like one of the most well done parts about the web platform. Not only from readability but it makes really good tradeoffs between verbosity and character count that makes CSS really really easy to read when written well. When writing CSS we have to describe complex relationships between elements, their state, and their relationship to other elements. With CSS, we are able to accomplish this with incredible ease.

The only parts I would change are:
- Make CSS cascade more and support nesting style blocks (like S[A/C]SS)
- Allow HTML Imports so we can write CSS in a <style> and not forced to use JS
- Take a another look at existing properties and reevaluate them.


You have clearly never tried to center something among several other divs.

.parent {
    display: flex;
    justify-content: center;
    align-items: center;
    text-align: center;

Flexbox is the best thing ever happened to CSS. ❤️

Agreed. Makes things so easy. Makes me hate the world a little for wading through the preflexbox world.

Late reply sorry, but nothing. Grid and Flex are to be used in tandem with one another.


Biggest CSS issue that has been kind of but not really fixed is the constant browser hacks needed.


The issue here is not character count, but predictability. The way(s) CSS works makes it so that there is lots of unintended and unexpected interactions.


I'm curious, what have you run into that was unpredictable?

Off the top of my head: adding margins that are ignored because other positioning takes precedence (I’m not talking about specificity here).

I’m curious that you don’t know what I mean with my previous remark, since unpredictability is pretty much a staple of CSS, so much so that it’s spawned jokes like mobile.twitter.com/thomasfuchs/sta....

I'd say the weirdest thing I've run into in that regard is margin collapse - like, why?

if you have a ul with some li and then on the li set display: inline then yes, margin will be ignored, but you can fix it by using inline-block. All inline elements ignore vertical margin. But this is more an "issue" with the expected vs actual result of the layout engine and not from CSS

Yeah, although what I was referring to is the way only the bigger margin is considered in some situations - you'd usually expect elements' margins to add up

The selector with the most specificity is the one that gets applied. If you have a element in an element in an element etc and they all have a margin then the margin does add up.

Margin collapse makes a lot of sense in textual content heavy documents. A typical example might be a li followed by a an h3 element, and generally the header would have an ample top margin, so there is no need to combine the trailing margin of the list with the leading margin of the header, it would simply lead to excessive white-space.

Useful for the kind of documents HTML was developed for, but not so useful once intricate visual designs and layouts enter the scene.

I agree with you, Meghan. Most of the hardship comes from not understanding how to do something in a simple way with CSS. As soon as we got rounded corners designers wanted something else we couldn't do easily. Haha


Had some fun with some of that unexpected behavior recently, apparently if you set overflow to auto on one axis (e.g. overflow-x), you can't set the other axis to visible. And it's in the spec, so it's definitely intentional 😆


Well, that’s the definition of design.


There is so much I would bin I can’t even begin. But top of the list, like way top would be no cascade. It’s such an impractical way of re-using code and over long periods in busy projects always leads to a knotted ball of string.

Way more practical is styling each component and encouraging component re-use.


Here's a post that doesn't answer this question but might provide some food for thought:


That was definitely a great read, and I definitely didn't try to top comment the thread by just saying CSS is great. I totally see where they're coming from, and tbh I don't think the web can do what Eric talked about. At least yet. Houdini has some good promise to make CSS a lot better and Custom Elements (with more browser support) might make HTML more expressive.


I think one of the biggest drawbacks and reason for the discomfort with why the web is bootstrap-lacking is that even though The Web was first made public until 1991, The Extensible Web only began in 2013. So there's this 20 year difference between how long we've been able to make web pages and how long we've made bootstrapping even remotely possible on the web. Thus, realistically the "modern" web is only ~5 years old. And if you look at just how much the web has changed only in those past 5 years it gives me a lot of hope for what we as a community will be able to do in the next 20 years now that everyone's on board :D


Here are the two things I'd change:


That was only useful back in the 90's when barely any websites had CSS, nowadays it's just not really very useful and hinders developers more than it helps them. Or, make disabling it as simple as this:

* {
    default-styles: none;

NOTE: Obviously keeping some default styling such as hiding <script> tags ;) ;)

2: Better ways to handle !important

You know when there's an element where you really want to say !important !important, because you've already needed to override something in a more specific selector? I think it would be good if you could treat importance like z-index, so you could give it a number - eg. !important(1) or !important(5).

What do you think about those ideas? I'd be interested to know.


About #1, do you know about * { all : unset; }? MDN Link


No I did not!!! Thanks for letting me know about that 😁


If you need !important that much, something is probably wrong and you need to refactor your CSS.


I don't use !important much, but having specific control over importance would make CSS more flexible


Considering the 'recent' grid and flexbox I think we're in a decent spot regarding CSS. I still do hate specificity in some cases (and !important can burn) but we're in a really great spot now compared with IE6 when I started developing.

Starting from fresh I only have three things on my list:

  1. Some sort of 'namespacing' so you can do 'BEM' on the html classes without beeing 'mycomponents__looks--like-candy'
  2. Any kind of nesting like SASS or your favourite pre-processor.
  3. Better handling of z-index and overflow

While I enjoy flexbox and especially grid, but I recently started using Elm and find the approach taken by the style-elements package very insightful. Layout and positioning is not really "style" it's architecture. Separating layout and style into different tools makes a lot of sense to me.


"CSS has a lot of issues." Which ones?

I feel that a lot of developers didn't learn CSS properly.


To me (as someone who naturally gravitates more towards the back-end of the stack) my biggest annoyance with CSS is that I find it all to be magic words. Nothing has a predictable effect, because how a property influences an element depends on what kind of element it is, what kind of elements it is a child of, what properties those parents have, etc etc etc. So when I set out to achieve a specific goal, it feels like I'm just trying random combinations of magic words on the element, its parents, its children and every other thing in the DOM until I find the "spell" that does what I want.


I used to center and align stuff using Bootstrap. Now that I've learned Flexbox and CSS Grid, I've found less use of Bootstrap. Though I don't use Grid that much. Flexbox is more useful for my projects.


I'm not a big Doctor Who fan but the TARDIS seems like it would be the best way to go, find the exact moment IE6 was conceived and just bloop it out of existence.

There would have been less need to cascade failure.

Other than that, ++Sass


Good question but to a large degree I feel like this question was asked of HTML and we got CSS. Maybe I'm too old but I like the cascading nature of CSS. It feels better than the old way of having tags and attributes everywhere to style things. I guess I'm in the, "CSS is awesome", camp.


To be honest, I've tried to do this for years - if we could throw away everything about CSS and come up with something similar, I'd still end up with something pretty close to CSS. It may look ugly, but it's fit-for-purpose.

You can check out some of the other styling language ideas from a time before CSS

When you compare some of the other ideas out there there are good feature ideas, but the syntax CSS has (which itself is based on X resources for unix programs) is simple, expressive, and flexible.

What I'm excited about now is that even if we want to invent a totally new styling language, we can use valid CSS + valid JavaScript to do it! CSS has "custom properties" (aka CSS variables) that allow us to define custom properties and set values for those properties anywhere in CSS, HTML, or via JavaScript. This is incredibly powerful.

There are also ways you can write CSS rules with custom selectors that get powered by JavaScript functions, as well as custom at-rules that get powered by JavaScript functions as well - so at every level (value, property, selector, stylesheet) you can customize valid CSS and extend it with small amounts of JavaScript logic to help it apply styles in situations, and to elements CSS alone can't quite reach.

The future of styling beyond CSS still looks like writing CSS, it's just designing and using a more declarative, higher-level styling language expressed in valid CSS syntax that is also powered by JavaScript functionality in addition to what CSS can do by itself.



CSS is missing programming language support. I just do everything with Glamorous nowadays because it lets you use all the power of JavaScript, without having to predefine every style condition in CSS.

Most people cringe at the sight of styles in JS, but that's because they think of inline styles. You can easily make a component-style.js file and done, you have separation of concerns.


You would not have to write CSS anymore. You would use a design editor that provided the design for all the browsers and just use the components as they are. The components would be just html elements with special tags on them so you can replace them with different tags if you wanted to make modifications. No CSS is needed.

All the logic for displaying the elements would be a machine representation compiled for the browser to parse and evaluate and would be handled by the design editor.

The design editor could be an external object or a plugin in the browser itself for quick prototyping.

Some low-code platform exists to offer similar capabilities but you still have to write CSS to handle edge cases and browser differences.


😕 Where's the fun in that? 😂😂
Personally I find CSS quite fun to use.


For me, I think inline styles can be a little more powerful.


if you are writing a typical website then sure another file for CSS makes sense, but for web apps where each component will be independent then Inline styles are good too.


I think now my only use for inline styles is styling stuff through JavaScript, which goes to the inline style


I feel like there should be a constraint: any replacement has to have the same feature set, including things like the ability to define user-agent stylesheets to overwrite the styling on any page.

This, sadly, gets lost with many of the modern approaches I see...


Probably one of the worst and best things with CSS is !important. It shouldn't be needed, but I'm glad it does sometimes.


i would make it easier for developers to contribute to css. right now css is split between whatwg/w3c/mozilla trying to tell browsers what to do, and then the browsers selectively agreeing/disagreeing.

this would involve the creation of some sort of annual or bimonthly convention, summit, or convention where developers get to hear talks from the authorities themselves, and possibly have a forum for discussing them. of course, as with developers being awkward, it would quickly turn into everyone staring at their macbook pro 2015s.

it doesnt hurt to go outside. css has value in its people, as with any other technology. if the people are valued by it, not just with it, css can change more in good faith, then


Remember when everything was essentially tables for layouts? 🙃 Or float based? Seeing people saying “centering is easy!” is both heartwarming and makes me feel like I’m 26 going on 80.


In my humble opinion, these would be some things I would include if we could take today’s knowledge and apply it to CSS from the get-go:

  1. Easier support for horizontal and vertical alignment. Grid and Flexbox weren’t always around, and when you had clients that INSISTED on using IE 7, even after it was officially declared to be dead, you are left with a lot of bad memories of shady hacks and cutting corners to make the browser act like it can think vertically, rather than horizontally. This is obviously much easier today, but it definitely wasn’t always something in everyone’s toolbox. Applying the knowledge of today, having this would be a no-brained

  2. Variables: another thing that may seem obvious, because this is becoming common place, css variables were not always a thing. Instead, I was taught to use SCSS for scalabitly and reusability. Going back, these dudes should be included in a redo

  3. Uniform spec interpretation: vendor prefixes are becoming less common, but there are still differences in how browsers render. Firefox handles margin and padding kind of differently without box-sizing, IE supported filters, then dropped them in a new version, and Opera Mobile apparently doesn’t support anything fun, according to caniuse. Even recently, I struggled with a task at work to add gradients around an image so it looks like it fades out of the background. Problem is, Safari interprets transparent gradients as black, where as Chrome and FF think of it as invisible (which is what I wanted). Neither are wrong intrinsically, but they’re different ways of reading the spec that quite honestly ruined my life for 3 days.

4 Nesting: it just makes sense


I like most of it, the cascade, the powerful selectors, the attributes. I would make only one change Instead of !important, I would have created pseudo selectors to increase or decrease specificy :+id.class.tag and :-id.class.tag, where id, class and tag are optional representations of the amount of increase and decrease. I also would have made the specificy of inline styles that of an id selector, so that it could be easily overwritten with :+1.0.1


I personally think that most of the issues commonly associated with CSS are the the result of misuse.
However there are a few things that could be improved:

  • Style leaks: there are several approaches to solve this but what we need is a native way.
  • User Agent stylesheets: 🔥🔥🔥

Just the way it does 😁
CSS is resilient and works


Make multiple properties have the same value:

h1 {
    margin-top, margin-bottom: 10px;

It wouldn't be a separate language. Bake it into js.

Also selectors being global by default is a flaw.


Bake it into js? NOPE. Separation of concerns is great. Put layout into the a new FLOW language, style into CSS, and actions into JS or web assembly.


You can separate concerns within the same language... look at all other "ui languages". Web dev is the only arena (correct me if I'm wrong) where you need 3 languages (html/css/js) to make a modern day component.

Mobile dev has fewer issues and has separation of concerns with no problem. If you used any preprocessor (less/sass) or template lang (handlebars, etc...) you proved my point. We wouldn't need that if js was the single language. This is one of the reasons why react is so popular.

Before I come off as someone who hates css, I use to only write html/css and hated js. Times have changed.


I'd like to fix style collisions in general. Probably get ret rid of global scopes and maybe an overhaul of specificity rules.


properties would be final.

once you assigned e.g. color:red, you are not allowed to override it in more specific selectors


Yeah I don't know if replacing it entirely with constraints is the right way to go, but being able to use them when you need to would be amazing.