- 273comments
- 183comments
- 2comments
- 219comments
- 198comments
- 41comments
- 124comments
- 9comments
- 3comments
- 113comments
- 15comments
- 15comments
- 130comments
- —discuss
- 25comments
- 45comments
- 119comments
- 329comments
- 171comments
- 396comments
- —discuss
- 16comments
- 31comments
- 234comments
- —discuss
- 2comments
- 78comments
- 69comments
- 12comments
- 61comments
Author here. Like a lot of people on HN, Dave Shea's CSS Zen Garden is how I learned CSS in the first place. Building stylesheets for that one HTML file taught me to think about CSS as a system, and it stuck. Nearly two decades later I got to build the design system for the new Firefox.com together with a team from Lincoln Loop, on a contract engagement with Mozilla using pure, modern CSS with no preprocessors (almost). In the end, we did run a few things through preprocessors for performance reasons, but the site's CSS is fully authored in pure CSS syntax.
Happy to answer questions about the layered CSS, the legacy browser fallback, how the tokens got exported from design files, or anything else you're curious about.
Is the code open source? It‘s not part of the protocol design system?
Correct, it's not part of Protocol. It's in this repo: https://github.com/mozmeao/springfield
I too was strongly inspired by the ideas that CSS Zen Garden was promoting, even if little to none of the real world work looked or worked like that.
Always weird to see what posts really bring out the negativity on HN, but pay that no mind. Some people can't stand others being happy (and some are just having a bad day...)
Be proud of what you accomplished and it must feel great to have contributed to something like Firefox!
The site is still live https://www.csszengarden.com, design selection here https://www.csszengarden.com/pages/alldesigns/ :).
Yes! Thank you. I linked to the site from my post as well. This is a great resource for anyone currently learning CSS.
This was the site shown to me to prove the separation of style and content. After that it was CSS-Tricks that I found myself frequently reading for answers to the things I was stuck on getting the layout to work.
Dave's old portfolio/blog is gone though: http://www.mezzoblue.com
The Bright Creative site is also AWOL.
Too bad neither is set up to redirect to the current https://daveshea.com/
Hope he's doing well. Guys like him, Doug Bowman, Jeffrey Zeldman... all a huge influence on teenaged-me hacking away on my little blog and thinking of going to SXSW someday.
EDIT: He's just left after being at Shopify for six years: https://lnkd.in/p/dcYBvQfq
I'm so tired of AI writing
It's a pity. Spending so much time to ship the CSS dream (clean, well-thought-out) and then using AI-talk to share it with the world.
Came here to say the same. I closed the window on the spot. I don't even care if I miss out on something.
I don't think these people realize how much damage they are doing to their reputations.
AI loves an adverb. Even the title feels AI to me.
"Honest" is not an adverb though? It's an adjective, the adverb would be "honestly".
Lol beat me to it
The below is a selected copy of the blog post, unbelievable that it didn't get unslopped:
And the entire analogy is incoherent.
And how all six people links are to LinkedIn, when most of them have websites. (Wouldn’t surprise me if some of the LinkedIn profiles were incorrect, but since you can no longer view anything on LinkedIn without having an account…)
I get the sentiment, but what if it was the difference between not getting the thoughts out there and being able to celebrate the fruits of their labor?
It's okay to exercise nuanced views (as well as choosing to say nothing.)
I'm sure I don't know what implausible hypothetical scenario to which this could refer.
The "difference" is that of spending five entire minutes proofreading the output. But this blog post, like the other 98%, is once more the tasteless fruit of very, very little labor. Whether you value it or not, whether you are even able to recognize it anymore, writing for an audience is a skill.
I (and probably GP) feel legitimate dismay when I read the output of someone who had an obvious opportunity to exercise good taste for the benefit of the people who actually take the time to read it.
The continued inconvenience to you of reading opinions like this is obviously extremely unfortunate.
Maybe they ARE lazy. Maybe they are unapologetically bullish on AI AND they did it just to spite the haters.
But it’s also possible that they have other, more important stuff going on in their life? Or that they find writing legitimately difficult?
I’m not excusing anything, or saying that you can’t feel dismay or even sickened.
I was merely proposing another perspective, that’s all.
Mostly, I have other, more important stuff going in my life. I did write a lot of this post myself, but used AI to flesh out some of the details because otherwise between my perfectionism and tendency to overthink blog posts, I would have never gotten around to publishing this.
Did you read it or, if English isn’t your first language, have someone who writes English well read it? I’ve had a lot of useful feedback on stuff I’ve written that way.
good good but are there really websites with 2 or more variously different designs for the same content, like Zen Garden did?
For some reason this is a thing on almost every private torrent tracker
Light theme and dark theme?
I'd say mobile and desktop (from same html) would be more like css zen garden, than light/dark.
Do they have the design system open source somewhere?
Considering the post ends with a link to the author's consultancy page, I'm guessing not.
If your team is rethinking its front-end foundation, get in touch.
Yes, you'll find it in this repo: https://github.com/mozmeao/springfield
Search for "flare" if you're looking for the CSS.
I tried some of the examples. The UX for seeing the different designs is terrible.
1. Go to site
2. Visually scan the page for the style controls
3. Click one
4. Everything changes
5. Visually scan the page for the style controls etc.
I dont want to visually scan every time I change the stlye.
https://csszengarden.com/216/
https://www.csszengarden.com/pages/alldesigns/
That non-CSS swappable layout is much better
The current firefox.com is incredibly simple, though. It's also pretty tablet/phone centric (everything is so large on Desktop).
Yeah... I was trying to figure out what he was pointing to, and was like.. oh, the Firefox website... went there and was pretty uninspired by the design, at least in desktop.
The CSS Zen Garden worked because everyone was using a single markup file. In real life, it doesn't really work.
Instead of thinking of separation of concerns as a religious principle, think about what benefits it actually yields in the case of HTML.
When the markup is completely separated from its styling, this means that the CSS rules need a way to be mapped onto the DOM. This involves specific elements, organized a specific way. If anything is out of place in the markup, the properties may not apply as intended. Similarly, if the CSS doesn't account for DOM structures, then you need modify rules or add new ones.
This way of mapping is an implicit structure. It's something extra that's required, beyond the source code that's there. It can be computed by the browser, but it's not explicitly stated for a developer looking at the code to understand.
Contrast this with Tailwind, Tachyons, etc., where nearly all the structure is explicit, less complex, discoverable, and out in the open.
Hidden structure looks nice to the developer. The code "looks" clean. But understanding a code base's implicit structure is ephemeral, and there's inherent technical debt that will need to be repaid later.
Tailwind looks ugly, but there are no hidden abstractions or structures in the code, beyond its easy conventions and a minimal config file.
And, yes, there are ways to make high-abstraction CSS work, say, with a component library of known DOM patterns. But at that point, why not build the styles into the components directly, a la Tailwind, instead of adding complexity?
I also dont think you can cleanly separate HTML and style. Technically you can separate them but doing so in practice just handicaps your design and ability to express things in HTML and CSS (Zen Garden prohibiting css 3+ in most cases).
People have this idea that they should be decoupled. You can do that if you really contort them. Naturally, they are coupled.
The structure of HTML puts constraints on the style. You have to know about the style when writing HTML, even in this Zen Garden project. You can't escape the style and HTML understanding each other. If you have a compelling reason to make a broad range of styles swappable then OK - but you are constraining the design space considerably and there is usually no reason for this complexity and reduction in expressiveness.
The document and the style depend on eachother. You see this with localization. You might need to change spacing or padding or breakpoints to accomodate different languages (text being different length in different languages).
This is why all UI paradigms tightly couple style and structure. QT, GTK, SwiftUI, Flutter, etc. Its the same with CSS/HTML except for people who delude themselves into thinking these were supposed to be loosely coupled.
Couple them how you want, HTML+CSS+JS is the system we've got, and it has to render well on mobile, tablet, and desktop. "deluded" and "supposed to be" are just word barriers that get in the way of doing shit that works.
The dream was that you could separate content from presentation and content could be adapted and reinterpreted by anyone using whatever tools they choose: User Agents that actually respected the users' vision, not the publishers'.
And accessibility.
The posts above claiming design and content are inextricably linked are probably blessed with all the standard senses.
As usual it's a bunch of folks talking past each other. Style and content together is a nice simplification when you want to conform to a particular vision. Separation is a nice abstraction if you want to enable multiple views of the same content (like a pdf version or a colorblind version or a vision impaired version) without having to change/control the content.
Like most things in this space, different tradeoffs make one better than the other in different scenarios. Taking a hard stance of "one true way" is probably missing some imagination or experience.
I think that's an original issue though.
Accessibility certainly requires a content-first approach, I agree with you there. If you view each part of an application solely in terms of what it looks like to a sighted user, it's easy to miss the subtleties of what the underlying structure ought to be.
But when you start layering on the presentation, you're still generally binding that presentation layer tightly to the current content. If you, for example, were to restructure the links in the Zen Garden footer for some reason, most of the styles would break because they're bound to the current structure of the content. The separation you talk about only exists in one direction.
By comparison, when I talk about separation and decoupling in terms of code, I mean that there is a single, explicit, and tight API that acts as a clear boundary between two units. By explicit, I mean that I've designed the units specifically so that they expose this API - this is almost exactly the opposite of what I should be doing with content and presentation, as you point out! By tight, I mean that the API is as minimal as I can make it - two units interacting via an API need to understand nothing of each other except that API, and that API is as small as possible.
None of that is happening when it comes to the interaction between content and presentation. The content is not explicit - I'm actively trying to write the content for the design, which precludes making it an explicit API surface. Nor is the content tight - again, I want to build the content without worrying about how I will layer the design on top, and I don't want to limit myself to just a small surface area here.
So to me, it makes sense to see the content and presentation as being coupled together, even if that coupling ideally goes in one direction only. And if two systems are coupled together, it makes sense to do that explicitly by co-locating those systems. Therefore, I tend to prefer systems like Tailwind that let me apply the presentation layer directly in the content layer. (In practice, I mostly prefer other tools that still let me write real CSS syntax, but Tailwind is pretty damn convenient a lot of the time.)
As an aside, talking about accessibility, I'd forgotten how inaccessible some of the Zen Garden designs are with their addition of extra content in the presentation layer alone.
Except HTML is already a presentational markup language. People lose track of this.
Consider: MathML has a presentational schema and a content schema. The whole point of presentational MathML is the intermingling of concerns. If you want interoperable data, you use content MathML.
Interoperability is a noble goal, but if you want HTML to do everything you're already barking up the wrong tree.
I’d previously understood the relationship to be that the document could remain unchanged, whilst the style (including visual structure/layout) changed. The simplest canonical example is that you can change the font, and all the words and meaning remain unchanged.
This is a really intereting case. Strictly I’d say it was a different document, if the language has changed. Really good translations change idiom, which can change meaning, and words, at the level of ohraes, although the intent would be preserved.
But I’m nit-picking, being contrarian, or exploring the edges of your comment.
Today, many web documents are app GUIs. Each field is essentially the same “thing” regardless the languge of its label. And in that sense, these are the same document, and your point holds.
I still think that if you can keep the markup as minimal as possible (reduce the div nesting to just what’s require to mark up logical sections of the content), then you could make a good stab at Zen-gardening your design. It’s very hard to do, however.
The real divide isn’t separation of concerns or implicit vs explicit structure. It’s whether you have staff that wants to think about markup as a data format expressing domain entities and/or generalization of presentation concerns.
Most developers and organizations don’t want to and they don’t want to think why or what doing so would buy them. They will probably find the tailwind local maxima enabling that reasonably comfortable and may even repeat the familiar mantra as has been done here. It especially pairs well with unreflective product management, which is the most popular kind of product management.
Verily, Tailwind is the assembly language of CSS. No structure, no semantics, no abstractions, only twiddling specific bits of visual representation.
To turn to it is to admit that your product lacks the structure and the design vision that allow to use some abstraction ("semantic classes"), and that all you can usually do is to patch some areas of it, disconnected from the rest, and unable to touch the rest (lest it goes down in flames). Assembly is definitely suitable for patching.
This is often the endgame of old large products that changed hands and directions many times, without much care.
This seems like a radical, unsupported conclusion to draw from a single frontend architecture decision.
Maybe there's an overwhelmingly positive case to be made for high-abstraction CSS with hidden structures. Outside those few select cases, we've known the tradeoffs for 25 years, and the added complexity is rarely worth it.
A right abstraction adds simplicity.
I enjoyed working with a semantically styled frontend project as recently as 2024. It also used a React component library, which helped insulate per-component styles, while sharing the common parts.
The dream of the CSS Zen Garden never really extended beyond itself. That wasn't the point. The point was to show people "look at all these pages that look nothing like each other: they're all the same HTML." That was the magic.
Keep in mind the context: you still largely needed structural HTML to effect the layout of your page. Almost everybody was building with HTML where the markup was critical to the appearance. Frontpage, Dreamweaver, and every other tool generated almost no CSS—or at least no good CSS.
This isn't my interpretation. From the site itself:
Oh yeah. The fucking CSS Zen Garden. It started the trend that led to utter degradation of any usability in modern web interfaces. The idea was to make HTML decoupled from presentation, but the feeble CSS of that time could not express any real layouts.
So very quickly web designers started to just hard-code pixel widths into the CSS, removing any dynamic sizing.
For example, on this very page: https://csszengarden.com/ the two-column layout is done by using two overlapping DIVs with a padding on the left one. You could do 3 columns by using the same trick with left padding.
Want to do four columns? Tough. You'd better redesign your whole HTML for that to work. Or just hardcode the sizes.
Of course, this quickly led to interfaces where buttons have text that is cut off. But with lots and lots of useless whitespace padding around them.
Umm... this is disingenuous. At the time, most websites were built using tables and were completely non-accessible. One of the reasons we moved to XHTML, semantic markup, CSS instead of Tables... was to enable accessibility.
No, it's not disingenious. Tables _were_ accessible and easily navigable, including by screen readers. They are also a very logical way of organizing content.
Again, nope. Try a design with 3-4 columns with the old CSS and see how it behaves. The order of elements in the HTML often differs from the visual order, confusing the readers.
It also breaks tab-navigation, even the https://csszengarden.com/ has the "View All Designs" button that is inaccessible.
Then add absolutely-positioned headers and other popular CSS patterns.
tables are a very logical way of organizing _tabular_ content. they're a terrible tool for layout, accessibility wise
No, they're not. They are _perfect_ for layouts, as they naturally support "move left/move right/up/down" directions.
To this day, this is STILL not working well for DIV-based designs.
I remember tables within tables within tables. Were those really accessible? The humans writing that html couldn't always keep it straight
Yes? What's problematic with nested tables? You just tab in and out of them.
Now try to navigate something like the WSJ/NYTimes home page.
If you remember the old days of good UI, there was this idea of "tab groups".
The postprocessor remains to fix some performance issue with @import? How does inlining imports fixes it? When I loaded the page, the font flickered.
Perhaps the article isn't about font import, those better be in html headers or simply stick to system fonts.
CSS imports create a dependency tree of stylesheets. You can't render anything until all of the styles are loaded and parsed. Imports create a blocking serial dependency.
I still remember being like 15 years old and seeing my website(s) land on CSS Zen Garden, CSS Thesis, etc... using Mint analytics (throwback! shoutout to @shauninman) to see all the inbound traffic. What a time to be alive.
I see no connection whatsoever with the CSS Zen Garden dream. CSS Zen Garden was about applying radically different styles to the same markup. This is about newer CSS features like Custom Properties, Flex and Grid making maintenance of a single stylesheet for a single completely typical website.
This is really great work, but I think a lot of folks on HN don't look kindly on the CSS Zen Garden or that era of CSS. They think the markup/stylesheet separation of concerns was a terrible idea, an idea that Tailwind exists to fix. (I think they're wrong, that separation of concerns is good, Tailwind is a hack, and get off my lawn while you're at it.)
The point is - now the modern vanilla web platform really can support first-class CSS-styled app components, from design systems, in production. This is really cool!
HTML and CSS were originally designed as document-oriented languages for digital document display; think MS Word or PDF with cross-document hyperlinks, and not an app SDK. The idea was simply "style your web site's titles, headings, paragraphs, footers, etc. in just one place; change it there and the whole site will update". That's the dream that CSS Zen Garden was meant to showcase. Interactivity and web apps came later, and CSS has taken a very long time to catch up.
But catch up it has, and I'll be glad when frontend web development sheds Tailwind and React in favor of the updated vanilla web platform; and I won't complain if we use super light web component libraries like Lit[0]).
[0] https://lit.dev
CSS Zen Garden in the era of agentic software development.... I would say we're at a point where if the entire site was developed in tables or CSS or a Rust based engine emitting wasm positioned elements - no one would care anymore.
Just give me a view and let me operate on things. We're going to have our "LCARS" moment soon. Where no one really needs to create a UI - the system creates one for you at time of use - based on what you're looking at and doing. We're no longer going to need UI/UX, everything will adapt to the person using things.
People say stuff like this and just... do you never think about accessibility? Have you never experienced stuff changing the UI under your feet and how incredibly annoying and flow-breaking it is?
It reminds me of people wanting Hollywood UIs that would obviously suck to actually use. Just completely ignoring the reality of using what you are suggesting.
People with AI Delusion Syndrome don’t really concern themselves with reality or other people. They’ve literally lost the plot in life under a grand delusion of corporate bootlicking.
You’re replying to a bot sadly.
On Chrome, Android, when I bring the menu up, and scroll to the bottom, the Newsletter menu item is cut off (I know it is there as I can see it on Firefox Android).
nails on a chalkboard. when will agents stop writing like this
Using slop constructs to convey intricacies about beautiful design is the purest form of anti-design.
That's where I stopped reading.
Remember this detail for later.
For me it was here. I have tried for months to get AI to write for me and have now fully embraced the fact that LLMs are not ready to do that for me. None of them. I’ve gone back to writing my own content and only using the LLM to brainstorm.