There are a bunch of comments commenting on other comments so I'm going to join in.
These comments are largely talking about how centring is hard from an implementation standpoint and can mean different things. Sure, but the whole "CSS centre meme" is just "How do I put one element dead centre within another". I can't point to data to back this up, but I'm pretty sure that's what 99% of people mean when they ask "Why is centring so hard?"
For those of us who grew up with GeoCities and AngelFire etc, the reason not being able to easily do this felt so ridiculous and why it was hard to buy that it's hard to implement is that we were able to do this no problem with HTML, and still can!
That puts the text "hello" squarely in the middle of a 600x600 yellow box [0]. This worked in the 90s. I know nothing about the implementation of CSS, but surely you can understand the annoyance of been handed a new tool, told it's better, yet wasn't able to do such a standard thing lots of people were already doing. It's a big reason I didn't listen to the dogma and kept up with table layouts into the late 2000s.
PS, maybe it's Stockholm Syndrome, but I love CSS FWIW.
[0] You actually don't even need the `valign` which surprised me, at least not in chromium... I'm not sure if that is a recent thing because I haven't actually written code like that in around 15 years.
You've been able to do exactly what you describe (i.e. putting "hello" squarely in the middle of a 600x600 yellow box) with CSS since its first incarnation, with e.g. the `valign` HTML attribute corresponding to the `vertical-align` CSS property on an `inline`, `inline-block` or `table-cell` box.
But then again, in my experience many of the people who wax poetic about just using tables and/or complain about how CSS is limiting and confusing...never actually put in much of an effort to learn how it works. At least the article's author is self-aware about this.
>> You've been able to do exactly what you describe (i.e. putting "hello" squarely in the middle of a 600x600 yellow box) with CSS since its first incarnation, with e.g. the `valign` HTML attribute corresponding to the `vertical-align` CSS property on an `inline`, `inline-block` or `table-cell` box.
Notably not a `block` element like a div or p - which is what most people are going to use as a container.
...the entire point of CSS is that you can style any element (almost) however you want, which includes changing how it's laid out by the browser.
Case in point - unlike with HTML alone where `valign` only works on table-related elements, you can set `vertical-align` with a compatible `display`/layout algorithm on a div, p, etc as you wish.
In 2000-2010 not only I was a fullstack developer, but also had to write CSS and HTML code for the websites I worked on often.
I remember that as an extremely painful experience. I can't remember all the details, but I remember that until flexbox and grid were implemented, everything CSS related was a pain. I vaguely remember that until 2008-2009 (maybe I'm off by 1-2 years) something trivial as rounded corners also was a pain.
I also remember that about 80% of the CSS pain was caused by Microsoft Intenet Explorer 6.0. It was complete garbage when dealing with CSS and Javascript (I was working with huge JS codebases at the time). But nobody in the web world could ignore it because it had a huge market share. I hate that browser and Microsoft with passion till this day. So many hours, days, weeks wasted to make everything work in a browser that wasn't updated by MS (why spend money when everybody is using it - they even fired the team that worked on it) and had a very "specific" implementation of Web standards.
You’re 100% correct in your recollections and the frustration of using CSS in those days contributed to me not wanting to keep working on the front end at all. And yes, F MSIE all the way to hell How many thousands of human lifetimes were spent trying to make that piece of crap work.
I still don’t know flexbox because I had given up by then. Thankfully I can have ChatGPT help me these days if it comes up, but thankfully FE is not really a part of my normal job.
You aren't wrong, but I don't think it was the case of not doing diligence. I welcomed CSS and started using it when it came out, though it was certainly painful back then (as siblings have pointed out) and I don't believe it would be a good faith argument to say otherwise. I was aware of `display: table-cell` but it always felt like a hack. I'm well of the irony in that statement, but if I'm going to call something a table cell, why not use a table? Regardless of how that sounds, that's how people think, especially 19 year old me! Furthermore, the name was far too technical. People who had little-to-no experience with table layout weren't going to search for "table" when wondering how to centre something.
In any event, the main reason I stuck with tables (and this of course was only in person projects) was because CSS didn't have a viable way to do layout until 2017.
Semantics aside, I still view table layouts as superior to floats and clearfixes. I was terrible about thinking about accessibility back then, of course, so I'm not sure of the impact there.
> You've been able to do exactly what you describe (i.e. putting "hello" squarely in the middle of a 600x600 yellow box) with CSS since its first incarnation, with e.g. the `valign` HTML attribute corresponding to the `vertical-align` CSS property on an `inline`, `inline-block` or `table-cell` box.
display: table-cell came along with CSS 2.0 (1998), and display: inline-block came along with CSS 2.1 (2011).
Unfortunately, there were huge amounts of CSS 2.0 that weren’t supported by Internet Explorer 6 (2001), and there was a five year gap until Internet Explorer 7 was released in 2006, which didn’t improve CSS support much (basically just support for a few selectors was added). It took until Internet Explorer 8 was released in 2009 before it supported the eleven-year-old CSS 2.0 properly. And after that point, web developers had to wait years before enough people had upgraded that it became possible to drop support for Internet Explorer 7 and below and start relying upon display: table-cell etc.
So even though CSS had display: table-cell in 1998, web developers couldn’t actually use it until after ~2011 or so. But still, that was 13 years ago now – time to stop beating the “Centring in CSS is difficult” drum.
Yeah - I was squishing together the current state of things with the original. Maybe a clearer way to phrase that would have been that CSS has been able to vertically align elements like `valign` on a td does given a compatible display from day one, with the compatible displays starting with `inline` and eventually including `inline-block` and `table-cell`.
Sure but we were told to move from tables to divs and "just use table-cell if you want vertical align" was not a great story nor was is discoverable to those had no prior context of table layout.
>n my experience many of the people who wax poetic about just using tables and/or complain about how CSS is limiting and confusing...never actually put in much of an effort to learn how it works.
I think the big issue in the 2000's was cross compatibility, and how seemingly simple elements on one web engine simply wouldn't work or look right in another. IE 6 was infamous for this. Tables were one of the few consistent elements, could be dynamically generated with ease, and simply feels easier to think about from a designer POV
I think that misalignment (no pun intended) is what lead to the huge snowball that is using frameworks upon frameworks to solve issues that some simple HTML/CSS should have solved. By the time the web engines evened out (and IE died) and CSS got better options, the damage was done. JS ruled the landscape.
in my experience many of the people who wax poetic about just using tables and/or complain about how CSS is limiting and confusing...never actually put in much of an effort to learn how it works
As somebody who was actually there back then your opinion reads as arrogant and condescending. Do you really think the entire industry just did it wrong for decades? That people did not frantically search for an easy solution for this silly problem they faced constantly? You'd do well to learn to be more humble.
Having also been in this industry for two decades, you haven’t noticed the amount of wrongheaded groupthink that leads the entire industry astray with flashy marketing for the latest ill-conceived toy for both management and engineers alike?
Yeah, this industry seems particularly prone to it. Do you not remember the 90s and early 00s, where most of the industry was utterly convinced you needed unreadably small body copy and splash pages? Or how about CSS in general? I must have heard “you can only use CSS for boxy designs” a thousand times before Dave Shea got fed up and created CSS Zen Garden to conclusively prove them wrong. The number of times I’ve seen the majority of the industry believe something completely dumb and then figure it out 5–10 years later is insane.
I have, but in this case it was truly necessary to do all the disgusting stuff like using Javascript to center stuff, because the content width had to be determined in hindsight. All suggestions in the article use features that came out in the late 2010s that solved this.
The entire industry did, in fact, do it wrong for decades.
The problem was never that CSS - a DSL that exposed more of the browser's layout and painting engine than HTML did - was somehow less capable than HTML alone at implementing the same design.
The actual problem was that the requirements of designing for the web changed very rapidly. Take GP's example of text centred in a 600px yellow square; again it's not that it was soooo hard to achieve that specific layout with CSS. It was that they were trying to accomplish layouts a lot more complicated than that, often without breaking those layouts down well enough to be expressed in terms of the available layout algorithms. The longterm solution to this was of course for browsers to provide implementations of more advanced algorithms (like flex and grid) to match more advanced layout requirements, but that process was dragged out for far too long and eventually landed in a far too fragmented manner - a massive mistake.
And on top of that, it's not as if the existing layout algorithms were particularly incapable; on the contrary, they are well documented and you can get very far with them alone. So many questions ("how do you center a div??", "where are my margins??", "why isn't z-index working??") have answers that are quite apparent if you're familiar with the spec and thus understand what the browser is actually doing when it lays out your elements. But how many devs even know that e.g. stacking contexts exist, to talk of when they are created?
It's not as if it's even their fault, it's yet another thing that basically the entire industry did wrong for years. How many web design/frontend tutorials really explain anything about the CSS spec or about how UI rendering works in general, as opposed to just shoving arcane styling incantations at learners to cram? It's a failing of the industry that even today you have to go quite a bit off the beaten path to actually learn CSS (and HTML itself, while we're at it). Imagine if we taught any other language in the same haphazard manner! Just because they aren't technically programming languages doesn't mean that they're not worth proper instruction.
If you look at any of the suggestions in the article, each uses at least one feature that came out in the late 2010s. Add a couple of years until initial bugs were fixed (which were there), and until browser adoption is high enough that older browsers can be ignored.
Back then I could whip up a dozen centering solutions, but not a single one that worked across all browsers in all cases. Tables just worked.
You're so wrong about the history that it makes me assume that you started webdev after this compatibility mess was fixed. It was truly disgusting. If anyone had found a solution back then everyone would have slurped it up immediately.
Ha, ya you're right. My memory on table-cell was fuzzy and didn't bother looking into before making some other responses. IE 6 and 7 had no support for it and Firefox had partial support until 2008. I'm just looking at caniuse, of course, I don't remember what Netscape supported.
I see overlapping text among a smorgasbord of other mistakes on the New York Times front page. Presumably they have the resources to hire top designers from this “industry”. No, I have no problem imagining that they did things wrong for decades.
I agree with most of your comment but my sibling commenter isn't wrong. I certainly think we all do do things wrong. I mean, someone put JavaScript on the server after all ;)
> corresponding to the `vertical-align` CSS property on an `inline`, `inline-block` or `table-cell` box.
I can't believe I didn't know about this until now.
Before flexbox and grid, I don't know how many times have I written `position: relative;` on parent, and `position: absolute; top: 50%; left: 50%; transform: translateX(-50%) translateY(-50%);` on the thing I wanted to put in the middle (vertically and horizontally).
Works in Firefox, too, without the "valign" attribute. Though I would not rely on auto-layout behavior in general since I've always seen subtle differences across the three major browsers whenever using tables without a fixed layout.
Nice article, the comments here are kind of amazing for a nominally technical audience. It is almost as if people have no idea how difficult automatic page layout and formatting is. There are literally PhD thesis topics on it[1]. And to expect that complexity to be abstracted away into some sort of simple "do-what-I-mean" expression? That just isn't going to happen.
Go look at Gwern Branwen's web site[2]. That is art. But the trick is decide how you want the site to look and then constraining your written material to be expressible in that style.
I've been looking at web page layout since 1995 when I joined a startup that was doing the "first magazine on the web about Golf!"[3] When the Zen Garden folks did their web site and started the 'A List Apart' mailing list which is now a website[4] it really helped me understand just what one was up against if you wanted to produce web content that rendered nicely on a wide variety of projections. And yes, the term projection is intentional because the function of going from semantic content to presentation on a screen or paper or other flat surface of finite size, is a mapping (or projection) from a native space into the rule set of the destination space. That rule set consists of both physical constraints (pixels per inch, total pixels horizonally and vertically, color capability) and software constraints (how much of the underlying capability can the browser software that is currently running express). Not to mention that every browser wants to do their own special thing.
So yes, CSS is a "hot mess" for people who decide one day "I'm going to build a web page from scratch." And yet, that mess is really just an abundance of choices rather than constraints on what you can do. The process is the same for everyone, find the tools that help you achieve the results you want and then package your material into a form that you can easily convert that into that look.
Hi Chuck,
Great points. FWIW, your tenure with web layout has me beat by about 3 years (1998). In all my relevant experience, the materials at "Every Layout"[1] remain unparallelled. Highly recommended.
Huh! Yes, yes, Hotdog[1] was what got me started with HTML. It came in a CD with a magazine that was about a month late to a town with no Internet in mid 90s.
The problem with css is almost entirely self inflicted, though. Yes, layout is hard. Why make it harder by aiming for the model we now have? Specifically, why aim for one major model that will fit all pages? Usually done with another sysipheon aim of automatic layout recalculated every page.
Combine this with the amusing goal of targeting any and every size of window. Why would anyone think that is doable?
Why don't we have completely separate styling systems for mobile devices, widescreen devices, vertical monitors, tablets, and billboard displays? Because that would be a bigger nightmare to deal with than CSS.
I like that when I visit a web page on my mobile phone, it loads regardless of whether or not the site owners hired someone to build a completely separate interface. And I like that when I design interfaces for phones and for large screens, I don't need to learn 2 languages to do it. And I like that when I stack two browser windows next to each other on a 1920x1080 monitor they resize and I can read both of them.
When people say that they don't want to worry about multiple screen sizes in their layout, they mean that they want an interface that works on one screen size and ignores everything else. And that would be a huge loss for accessibility and innovation if the web pushed developers in that direction. The reason it would be simpler for people to work with a more targeted language is because they wouldn't build the other interfaces at all. They'd learn one language, target one device that their most predominant customers used, and then we'd have a mobile Internet and a desktop Internet and they'd be separate things with no expectations that sites would work on both devices.
----
And I think this sort of gets to the complaints about complexity in general, because the complexity of this UI design is reflecting a reality that good interfaces are adaptable and people have multi-faceted needs from their software. Even on desktop, people use different screen resolutions, they scale fonts, they mess with layout. And there's this subtle idea behind complaints about complexity that when you dig into it is not actually "why do I have to target so many devices" but is really "why do people use so many devices? Why isn't the world more uniform, why on earth are people changing their screen resolutions, what's wrong with them? Why can't they just decide on a device and stick with it?"
But good UX design is about designing for the real world, not for a hypothetical standardized human, and in the same way that cars need adjustable seats and can't just say "well on average everyone is this height and width", good UX acknowledges and responds to the idea that software and content are delivered in multiple contexts.
Of course that's a balancing act, it does make interface design more complicated, and it's not something we can do perfectly. But it is a balancing act, it's not a problem we can solve by saying "heck it, everyone needs to stop buying HDPI laptops." I mean, we're not all Linux developers, we can't all just pretend that touchscreens don't exist ;)
The same exact complaints show up with the extensible web and with progressive fallbacks in general. It is real annoying to build software that degrades nicely depending on what hardware support someone has and what features they've turned on and off in their browser. But it's also a better way to build software that better reflects how software is used in the real world by real people.
A problem is that none of the arguments you are offering are bad, logically. It feels very appealing to think you can build a system that would solve "laying out text" for all time.
For me, what they lack is evidence that stands any stronger than video games. Specifically, I've played different games on various screen sizes and orientations that largely work as you would expect. They are not perfect, of course, but they work better than most web pages seem to. And they do that, largely, without the same reliance on something like CSS that web pages need.
More, it isn't like we weren't laying out billboard displays long before the web came to be. Nor is it realistic that billboards have at all the same concerns that a pocket sized phone will have. At large, you shouldn't even use the same fonts between those options. Heck, taken farther, a billboard can hold a slogan, that is about it.
This would be the same as if you tried to use HTML/CSS to make a poster for a movie. Which, sure, you can make a bit of an effort with it. I just don't see it being any better than letting a designer or probably an automated system layout several standard sizes with the standard type in the standard locations.
Pulling it in, I'll be delighted to get proven wrong and find that we have converged to a great abstraction for laying out content. I, of course, do not /know/ that it can't be done. I do pull my hair out at the amount of effort people will go to in order to have the system layout a set of divs, when most designs could probably have done a lot of that math up front and worked with far fewer nested elements than we seem to typically see.
Good layout is simple. But can it be flexibly done without a good sense for abstraction and math? No. I mean, just look at the hot shit that is Tailwind, that's what people come up with while fighting abstraction.
CSS was a mess, but together with React is now the best language for building GUIs. You can build up any abstraction you want in React, even
In fact, that's what I did over the weekend for my Electron desktop application, writing my own little library of components, including tabs and treeviews, using only CSS and React and no third-party libraries apart from that.
You already have been proven wrong. You are just not ready to accept the proof.
Where was I proven wrong? Does that example compile down to something that isn't several dozen nested divs using react?
As stated in my post, happy to be proven wrong if that is the case. I am fairly far removed from a lot of this nowadays, and it would not be the first time that the world moved on without my noticing rapidly. (My favorite example of this is just how good battery technology has gotten. A decade ago, a battery powered lawn mower was a laughable idea. )
> For me, what they lack is evidence that stands any stronger than video games. Specifically, I've played different games on various screen sizes and orientations that largely work as you would expect. They are not perfect, of course, but they work better than most web pages seem to. And they do that, largely, without the same reliance on something like CSS that web pages need.
Video game interfaces are a ton of work, and porting between different control schemes and devices is a ton of work and that's why a lot of games don't do it. Look at the work required to handle devices like the Steam deck and the amount of work Valve has put into trying to make mouse-controlled games usable on the device. It's not easy, it's significantly more work than building for the web.
Of course it's helped by the fact that Valve does have general abstracted concepts of input that games can hook into that are shared between different controllers. Modern games don't do input per-controller, they use abstraction libraries like SDL that are designed to allow them handle a lot of different input schemes with a single codebase. And even that is a crapshoot, if you're playing indie titles on a PC you are going to be rolling the dice on whether Xbox controllers and Dualshock controllers are supported or whether only one of them works. I like that when I use the web, even weird keyboards still type into webpages. We still don't have a standardized way to do controller rebindings in games -- Valve's Steam Input works by emulating a second controller for games that aren't using Valve input APIs and pretending to press the buttons the game expects to see.
So even with all of the advantages of cross-platform frameworks, the games industry still doesn't have a great track record for building games across devices. You're pointing out that there exist games that do build separate interfaces for different devices. Sure. And agreed, when devs put in the extra work, you can have great results that are far better than the average website.
The difference is that basically every website works on my phone. Games aren't even close to that level of compatibility and most teams don't have the resources or time to put in the work to support every device and control scheme. The world of games is exactly what we don't want on the web, because if having a website interface for mobile and for desktop means that everyone needs to design and program two separate interfaces using two separate languages, we will have about the same number of websites on both mobile and desktop as we have games on both mobile and desktop -- ie, very, very few of them.
It's a blessing for compatibility that engines like Unity and Godot allow targeting multiple platforms and form-factors with a single codebase. I wish I had to work less to get that same level of abstraction in my video game interfaces.
And all of this is before we even get into all of the other problems of game interfaces -- the all-too-common lack of ability to do anything with text sizes, the lack of accessibility controls, the inability to resize windows in or out of games. My goodness do I not want my web pages to act like video games, that would be a miserable experience. It's not uncommon for me to see video games that literally won't allow changing resolutions without restarting the game. Imagine if your browser forced you to close and re-open the webpage in order to resize your window.
----
> It feels very appealing to think you can build a system that would solve "laying out text" for all time.
We have not built a system that can solve laying out text for all time. That system doesn't exist and can't be built, it is impossible. That's exactly what we're saying: we are trying to build a system that does the best possible job of universally solving that problem, but it's incredibly difficult and necessarily results in complexity and tradeoffs.
But the only belief more naive than thinking we can have one universal layout system for everything is the belief that we don't need a universal layout system and that it's possible to reduce user needs into a finite list of use-cases that can be individually supported. That reduction isn't possible, the use-cases for end users are arbitrarily large and constantly growing and it is not possible for anyone to sit down and build a single list of formats that need to be supported on the web. That's just fantasy, people are too diverse.
I use both a 1920x1080 monitor and a 3840x2160 touchscreen monitor hooked up to the same computer. Every single website I visit handles both, fluidly, when switching windows between them. A lot of games have no idea what to do and a nontrivial number of native apps struggle with it as well. But the web works, and then people show up like, "well, I shouldn't need support that." Well, I'm glad you're forced to use CSS then because I know what the web would look like if you weren't forced to.
Yes, getting an interface to work is a ton of work. If you felt that I was claiming there are easier ways that are not a lot of work, my apologies. I did not intend it that way.
My point would be more that the work Valve has put into the Deck has enabled far more than the work the standards committees have done with CSS. You can correctly argue these are solving different problems. But my assertion is that I have seen more impressive content layout and interactions from the Steam Deck than I have really with the web. I'm curious what you'd offer as the reasoning there?
You seem to be taking it that the games industry isn't very cross device focused. But, that is missing that current gen games are usually at the absolute bleeding edge of what the absolute best devices are even capable of. It is not at all surprising that those do not work cross device that well.
You are, of course, correct that there are some generic engines that allow better cross device development now than have existed in the past. Do any of them use something like CSS for laying out a menu screen? If not, why not? How about the inventory or character creation screens of games? Would you think those should be designed in the same way that something like a character builder webpage would use?
I also have large monitors, and it is ridiculously amusing how many websites do not work well when I have my windows tiling. My favorite is just two windows side by side on the main monitor, but parts of the menu of many sites will not load due to confusion over what my window width should be. To my absolute annoyance, I have found this will often not be consistent between browsers.
Now, is it fair that "bugs mean the entire thing is nonsense?" No. And I do apologize that I can see how my post read that way. Realistically, the amount of manpower that has gone into CSS has landed on something that is quite capable. But we gave up on "user stylesheets" ages ago. And the cascading nature of how things interact is almost certainly not well understood by a large portion of the practitioners. My annoyance is far less on what is capable with CSS nowadays, and much more annoyed at the rube goldberg machine that is how the vast majority of websites are layed out.
All that said, there’s still no way to just make a layout and it be compatible with all screens. You still have to do mobile, tablet, pc modes. (And there’s enough sites that are pc-only or mobile-only.) You still have to make 2-3 layouts, but in one source. Well, what’s the argument against gp’s point again? Sorry, but this statusquoism and rationales that essentially change nothing drive me mad sometimes.
The web and CSS make it reasonably feasible to support most of those layouts well with at most 2-3 layouts in one source.
My point is that it is impossible to get the same level of results doing fully device-specific and domain-specific languages without considerably more work -- more work than would ever be undertaken by the majority of people building for the web today. If you want an indie web, if you want websites that aren't just built by corporations and professional development teams, HTML and CSS is what enables those indie websites to exist and to work across every device you own.
The proposal of using interface-specific languages and abstractions wouldn't work. Doesn't work, in fact, when you look the places where it's being attempted as a UI strategy.
Like democracy, universal design abstractions such as CSS are the worst way to design interfaces that work in the real world for real users across diverse devices and configurations -- except for every other way. If we asked devs to actually use separate domain languages for mobile devices and desktop devices, the web would be a lot smaller and a lot more limited and a lot less flexible and a lot worse for end users.
You can just make one layout that works totally fine on all screens. Problem is you're basically just making a mobile layout and scaling it up so it'll be shit on larger screens
Heavy, heavy disagree. Tiny example, but just because it's on my mind that I need to rewrite it soon: https://danshumway.com/resume.
This is almost per-pixel identical to the resume layout that I used to have laid out in a normal editor exclusively for PDF and print, but without making a single concession with my normal print layout, it also seamlessly collapses into single columns perfectly fine on mobile devices, doesn't require Javascript, and the entire layout is ~160 lines of CSS.
And to be clear, this is bad CSS. This is not CSS I'm proud of, this is CSS I wrote ages ago that I'm going to be tackling very differently whenever I get around to rewriting the resume. But it's still perfectly reasonable to have document layouts that feel like normal print layouts that collapse down well on mobile devices.
And sure, this is a trivial example, but I think the same principles scale up. I don't believe that applications are different. Most application UIs can be represented as trees, and trees are well-suited for multiple-column layouts and collapsing rules based on width. That many sites work on mobile and don't scale up is a result of developers designing primarily for small screens and not thinking about large screens at all.
Even here, this resume should handle wider screens better, it doesn't take enough advantage of available width. But the other thing about the web compared to native is that with native toolkits if a developer designs for mobile and then doesn't think about desktop, the app just doesn't work on desktop at all and you can't use it.
On the web, if a developer designs for mobile and doesn't think about desktop, at least the interface is still usable on desktop. And vice-versa: there are desktop interfaces on the web that are miserable on mobile devices. But they work. None of the alternatives that people are proposing have that advantage.
I recently spent a few days hacking flow-based layouts and pages into a flex-box-based layout system, and I can confirm that layout and formatting is hard. Figuring out how and when to shift elements to the next page, especially when you have a table with columns containing text and objects of different sizes, is a challenging exercise in picking good heuristics.
How to make sure it’s not self-imposed problem and not in-there research? In my experience, I never thought “layout hard” before joining web dev. And yes, we had scrollable resizeable windows back then. I hear that the status quo is the best way from everywhere. But then I remember my custom appkit, gtk controls and even my own (unreleased) lua-based toolkit, where centering content or wrapping and aligning elements wasn’t an effing deal, and wonder what am I missing.
And yet, that mess is really just an abundance of choices rather than constraints on what you can do
Also known as TMTOWTDI, long forgotten and buried for good.
What you wrote at the beginning is: "It is almost as if people have no idea how difficult automatic page layout and formatting is."
However, the rest of your text is: "look how difficult it is on the web, and look at all these amazing things people do despite this being so difficult on the web".
All the complexity on the web is 100% self-inflicted. It started as a simple system to display a couple of paragraphs of text accompanied by a couple of images in one rendering pass. That's it. And then it grew haphazardly, with no roadmap or plan, through a series of hacks bolted on top to give it more capabilities.
We could do the "near-impossible PhD-level stuff" in 1970s [1]. The Mother of All Demos from 1968 arguably had more capabilities than the web even today: https://www.youtube.com/watch?v=yJDv-zdhzMY
For some reason you took "it's difficult to do on the web" as an inviolable axiom of "it's difficult to do, period". The entirety of "impossible layout on the web" is about as complex as what WordStar could handle in mid-1990s
> Nice article, the comments here are kind of amazing for a nominally technical audience. It is almost as if people have no idea how difficult automatic page layout and formatting is.
It's nuts that this is true after so many years of HTML and CSS.... software sucks any way you look at it.
Why we still can't get a bug-free OS is beyond me.
Great article. Love the interactivity. One thing that really helped me understand CSS positioning (and centering) years ago was reading about the box model. Understanding the box model helps you determine flow within the DOM. The `display` and `position` CSS properties are also fundamental to learning about positioning. MDN has great articles on these!
The "how to center" is a classic question for CSS. The fact that it is not blindingly obvious is indicative. CSS is a miasma. A kitchen sink collection of stuff designed, not by one committee, but by several committees simultaneously.
They have long since given up on actual releases. Instead, the current state of CSS is a collection of states of individual modules, all of which are under continual change.
This is not how you develop software. Or anything, really.
I both agree and disagree. This is just how how larger groups of humans work, slowly and chaotically. Herding cats. It's also not clear where we're collectively going with the web and all of its technologies, so there will be lots of failed experiments and disagreements. There's a much broader scope here than a well planned engineering project with a clear goal and path.
Regarding centering - eh. I think this question comes up so often because it is simply the natural question to ask.
What are the first things someone learning how to make a web page visually look how they want is going to learn? Probably text size, color, background color, and alignment. All of these except alignment translate pretty much 1:1 from pre-css days. Color on the body tag, size/color on the tag surrounding the text you want styled. Or, css color/font-size property on the surrounding tag, or background-color on the body. All very similar.
Prior to CSS, if you wanted content centered, you just slapped it in a <center> tag. Didn't matter if it was text or html objects like divs, tables, buttons, etc. IMO too many people expected CSS to work like what we had in the past - no distinction between inline and block objects - and when CSS didn't fit their pre-conceived mental model, these questions came up. And centering just happens to fit in the niche of being basic enough that beginners want to know about it.
According to this, most CSS issues boil down to “you don’t know CSS and have a wrong mental model”. Well, if most people who enter it have a wrong mental model, it’s not their fault that the “correct” model is absolutely alchemic in its nature.
"Software that doesn't conform to my pre-conceived mental model is poorly made".
No, what you said is not the point of what I wrote. "Wrong" in the sense that it's an imported assumption that has nothing to do with the technology at hand. And don't see anything wrong with that. Having an open mind when leaning something new is important.
Agree on the statement, disagree on the sentiment. You only have that much capacity. If people have preconceptions that clearly correspond to tasks/solutions they have in mind, it's absolutely unwise to ignore this free part in your design and design something that makes it completely irrelevant.
It's taken decades for CSS to come up with anything as workable as putting content in a centered table. Yet during that same period, use of tables was shamed for layout.
HTML Tables don't wrap on mobile displays. If you build your site with tables for layout, it will probably have horizontal scrollbar(s) and you'll need to rewrite it for mobile; so a CSS framework is usually a safer choice for less layout rework, unless it's a data table.
I can't tell if you're serious or not, but tables are an absolute mess to handle layout.
Email clients handle the CSS code to varying degree of support when wrapping responsively so even if you think your layout looks good with wrapping it's because CSS influenced the layout in whatever particular client you are using. If client doesn't support CSS or the CSS properties you've used: jank email. Email unfortunately requires you to code like 1999 in this regard.
Browsers having CSS as standard and relatively up to date is a very good thing. Tables that aren't representation of tabular data are a mess.
It's actually really simple to use tables to handle layout once you're used to it, like I said, it's what we had to do with email for a long time.
It does take some css, of course, but only width and max-width. That's basically it to turn a table responsive. Add the dir attribute and you can also control the stacking order of elements.
My point being, working in a constrained way (like with email) allows you to make the most of the properties available to you, and it turns out tables for layout (semantic issues aside) are easy to grok and work predictably and reliably.
Tailwind is anti-semantic. It's a less verbose version of the style="..." attribute. It's like people just gave up on the idea of element classes altogether.
And good web development is harder because still too often web design is done by designers who don't have enough understanding of web development.
My gawd, just earlier today I saw a help wanted that said "...PSD to pixel perfect..." I would have laughed if I wasn't already crying. "This is my life" and then the tears kicked in.
- CSS back then was wildly, W I L D L Y unsuitable for what people where trying to do (I think it boils down to it being designed to format documents, not web apps or on-screen layouts)
I never had to spend time in the old days debugging centering, I just used <center> and tables and it worked! Then I could spend the time saved by adding <font> tags everywhere and making images for my rounded corners.
Like many others in this thread, I think CSS is an absolute catastrophe. For me, the key issue is that style directives are constantly either 1) clobbering each other or 2) silently failing to have any effect.
It’s interesting to note what we can learn from the use of tables for layout.
Tables have border, padding, and a cell-based layout arranged in columns. At a stretch they and their cells need white space around them, ie “margins”.
Tables are a long established element in layout.
And it turns out tables actually contain this wisdom, that all elements should work a bit like table cells: they should follow the box model (fairly obvious to anyone working in typography) but also that they should be arranged in columns (much less obvious and the foundation of grid layout and bootstrap).
It’s a natural evolution to come across some layout problems, solve them with tables (as any typesetting system is likely to support tables), and then realise that you need a new facility that has some of the properties of tables but not all of them.
CSS has discovered this. Oddly, some systems have not - for example, word processors have not unified elements under a box model etc.
Wow, all these "why is this still so hard" complaints feel so off-base to me! As the article says, Flexbox straightforwardly solves centering in every simple case. When it doesn't solve it for you, you're doing something more complex than _just_ centering, so it's inappropriate to expect that the implementation will be dead simple.
It's because for those of us who have been around a while, it was hard before flexbox, then it continued to be hard when flexbox didn't have 100% support and we kept having to track browser usage and deciding whether we could use flexbox or not, and now there are just so many ways to do it that it's an overwhelming amount of complexity. You're often not building something from scratch but rather updating old code, and having to figure out why the CSS was done in one particular way for what reason, and what might break in which edge cases, and whether you should/can update it or not, and which one of several solutions when it's not simple.
If you look through the article, none of it is intuitive. Even with flexbox, the horizontal setting is called "justify-content" while the vertical is "align-items". I personally simply can't remember what a lot of CSS properties are called anymore -- there are so many now, and the names are so arbitrary. (Even hyphens are maddening -- why is it "white-space: nowrap" instead of "whitespace: no-wrap"?)
At the end of the day, there's a big difference between:
1) Simple, intuitive, reliable building blocks that let you build complex solutions out of simple parts
2) Convoluted, partially-overlapping, constantly-needing-to-Google building blocks that let you build complex solutions out of complex sets of parts
A language like Go or Python is #1. A language like CSS is #2.
justify/align and content/items have specific meanings and the 'intuitive' naming most people expect - horizontal, vertical - completely betrays the powerful elegance of the flex model.
justify always refers to the main axis of a box model element, and align refers to the cross axis. by default, flex-direction is set to row, so justify often means horizontal and align often means vertical.
until it doesn't of course. when we switch tbe flex direction from row to column which is a very handy trick for easy responsive on small devices, the main axis is now vertical and the cross axis is now horizontal.
There's just always going to be confusion because both pairs are basically synonyms: justify and align; items and content (what's the content of a flex container? flex items).
I buy the value in having them be main-axis and cross-axis properties rather than horizontal/vertical by name, but I don't know that I buy the justify/align distinction to represent that as being meaningful or memorable.
It's very easy to remember all this, if you can just memorize the acronym AMIJAMITCOE ("Ami-Jamit-Coe"): A monad is just a monoid in the category of endofunctors.
My mnemonic for justify vs align is to picture the "justification" options in a word processor. you can justify content to the side of the primary axis when it's written right-to-left. so, justify is main axis, and the one that's not justify (align) is the one that's not main axis (cross axis).
What I don't remember ever is does flex default to row or column; so I write a lot of unnecessary `flex-direction:row;` but if you are fine with always specifying a direction, the default doesnt really matter either.
Its similar with changing "left" and "right" to be "start" and "end" because we built in a lot of assumptions with LTR languages that don't apply to RTL.
They could leave the elegance by calling it align-x and align-y and turn-right. Or make them independent of “direction”. Or allow both ways so people could use them in corresponding situations. Instead they chose to satisfy the most useless case of a pseudo-turnable container, which makes zero sense.
Flexbox is what I always wanted CSS to be, but for a long time wasn't, until one day it was, and we could drop support for browsers without it. Feels almost end-game for CSS.
There was a reason people loved Bootstrap's grid system so much, Flexbox does all that and more straight in your browser. Cannot over-express how much I love it.
I too love flex but your comment is a bit off. bootstraps grid system is a fake grid modelled on flex. building layout grids with flex in general is definitely not endgame, its actually very tedious and unreliable.
css grid however is perfect for this and represents one of the most underutilised powerhouses in the css kit
I haven't really found much use for CSS grid. It's fine if you're doing a strict grid, but it seems to have all the limitations of a literal table to me. You can't put two objects in one cell. It has no way to wrap. It just isn't responsive at all. You can't make a sidebar with grid. You basically have to just write multiple styles and switch between them with media queries.
In practice, I find making a fake grid out of flex boxes much more usable. About the only thing grid does easier is letting you place a footer correctly. (Although you gotta make sure you count elements correctly because if you insert an extra element before your grid's footer, it will be rendered after.)
Yep, there was a time when it was moderately difficult in certain cases, but it's been a solved problem for the last decade or so. You still hear people acting like centering DIVs in CSS is something like a cross between alchemy and rocket science, which is perhaps indicative that people don't follow CSS with the same attention as other parts of web development—yet they still feel comfortable making these statements!
This is super true. Whenever I find myself struggling w/ flexbox it helps to take a step back and solve it one step at a time. Working my way out-to-in, parent-to-child element, and it becomes so much easier. There's a skill to it, but it's basically just elastic algebra.
`<div centered>` is the same thing as `<div read-my-mind>` since there are so many variations of what you might want. And once you generalize the solution around those variations, you have something like flexbox.
For example, what does `<div centered>` do when there are two children? And what do you do when you want the opposite behavior?
> For example, what does `<div centered>` do when there are two children?
Force me to use the advanced option or to wrap the two children in a centered container?
I am certain this wouldn't cover many, many edge cases, but they are called edge cases for a reason.
Edit: People can complain about tables etc being the wrong approach all they like, in the end those examples are only proof that the proper way could be at least as simple for the average case.
It takes the sum of two children’s intrinsic sizes with a margin between them and centers the bounding rectangle as if it was a single object. Flexbox does the same, and I didn’t even think of it while writing the previous sentence. But somehow you didn’t have the same read-my-mind complaint about flexbox.
Because flexbox picks a default when you give it no configuration but it lets you configure it, so it doesn't need to read my mind and it can be wrong every time.
Now show me how you change spacing/centering behavior with the proposed <div centered> and you'll finish out the point of my previous comment.
Contrast the need to use flexbox with the ability to just specify bold or italic style, or for that matter use <b> or <i> tags. You basically get what you want every time. Nobody has to argue about whether the italicization is the right slope, and while you can select a particular font weight if you want to, bold is bold and is fine for basically everybody.
Hell, I would argue that centered table contents gave people what they wanted most of the time, and if it didn't you could fix it by adding more tables. Tables composed quite nicely and gave you something if you just laid down a table and some TDs.
Flexbox doesn't scale down the way tables do. I have to have both the container and elements inside that container and think about the ramifications of all of that. It's minimum level of complexity is a higher floor than most of the HTML specification.
That's not to say it's the wrong solution for the problem; I think a case can be made that the problem itself is irreducibly complicated. But that's why I say it feels like using an elephant gun relative to much of the rest of the spec; I have a "read my mind" option for bold and italic.
Do you want your div centered vertically? horizontally? both? Relative to what? the window? the container? These are just some of the common questions that are asked every time you need to centre something. A simple attribute cannot work.
I don't think I ever wanted things centered relative to anything other than the direct parent, no matter what that is. I am sure some cases need it differently, but I don't see why those should prevent intuitive defaults.
> A simple attribute cannot work
Can it not work at all or just if one insists on covering 100% of cases on the internet with a single comprehensive option? I'm not an expert in this so I'm not sure, but I've seen a lot of software center items automatically and it worked quite well because they didn't assume by default that I might push the boundaries of design to new heights at any moment.
grid and flex were very nice semi malleable additions, they really lift the legwork and in the 10% of complex cases you can spend your brain time with joy
Frankly, I would probably just use flexbox for any centering use case. It's simple, fast, and straightforward. As far as I understand, flexbox has a pretty well optimized implementation, so it's not really a performance concern to just throw in another flexbox. Why would you not use it?
I often don't understand what flexbox is doing when what's inside is itself complicated. I think that fits-content might help, because I haven't tried that.
I honestly don't understand why people would find Flexbox hard to understand. Yes, there's a little bit of new terminology, and there's a reason for that. There is a main axis whose orientation can be changed, which affects the interpretation of other properties you use, but most of the time, you won't change it.
I don't know about that. I'm not a UI person, yet I feel like I have a pretty good grasp on CSS. I actually like it. So when I'm sick and tired of programming things, I've found myself recreating the layout of some site I liked by writing the CSS/HTML by hand. You learn pretty quick that way. It's a shame that all UI is now only allowed to be done with some library or other.
I was referring to the observation that getting a Ph.D. usually requires wasting 3-6 years of your life doing something tedious and useless (source: my time getting a Ph.D. in computer science)
The problem is it was easy to center things with tables. Then once CSS was popular, that was the "wrong way" to handle layout. I find CSS the worst part of web development, and it's what turned me off of front-end work. Glad flexbox has finally made things simpler again.
I like CSS but I'd agree that tables got a bad rap. They were very easy, very effective, and only occasionally turned into a nightmare of nesting that was impossible to maintain
My problem with CSS now is that it's gotten too bloated to the point where it's introducing privacy and security risks. I really want an add-on that restricts CSS by default to only a sane subset of features.
They're also awful to maintain over time, if you're needing to add new elements and require new columns or cells (and then there's the need to nest tables when using them for layout).
I attempted to fix a website by adding closing table cells that were missing, and finally got the page to validate as XHTML Strict and Transitional. The entire layout shifted, and was completely based on those missing end tags.
Not the point. The point was that some people still think there is nothing wrong with using tables for layout. The guy who started it all would like to have a word with them.
Centering like this can make your element blurry on non-hidpi screens, because transforms are not snapped to the pixel grid like layout properties, and 50% can sometimes land on a subpixel
Seems like they managed to kinda fix it in some cases, but some of the reproductions provided in the issue tracker are still working, so I would be careful
I was kinda surprised that this method wasn't included in the article. While I use flexbox for centering most of the time, I still find myself resorting to the absolute positioning way sometimes.
Really nice tutorial. On a related note, can somebody tell me how to make a nested div overflow (scrollbar) without specifying the height of all the parents divs?
<div style="height: 100vh" id="one">
<div id="two">
<div id="three" style="overflow: auto">
I need
</div>
</div>
</div>
This is a rather simple example, but sometimes the divs are nested really deep inside and unless I specify all the childs divs as 100% height there is no scrollbar. This causes me a lot of issues.
I solved this a while ago. If memory serves me correctly, you need to set overflow none on div two. Of this doesn't work, let me know and I can check my codebase next time it's convenient.
I'm too lazy/busy to confirm/test this, but providing a new stacking context on the parent by either by adding a position: relative (or whatever) or adding transform: translateZ(0) (or X or Y) avoids the need to specify every parent's height. Someone less lazy than me could confirm :)
I haven’t found a way to center a div with a background color containing text such that when the text wraps the overflow space (tried to expand horizontally to this size but ultimately needed to wrap) ends up being reclaimed (not background colored). Without forcing text to wrap at a certain width if it might still fit on one line.
How come the desktop UI frameworks have this issue solved for decades and CSS/Web still screwing around with mishmash of solutions? I guess, people were afraid to break stuff and kept piling garbage on top of garbage. I think it is safe to assume that the web layout systems suck.
Reasoning about the qualities of a system without having a competing party doesn’t do good in politics (dictatorship), in business (monopoly), in browsers (ie). But it’s different when it comes to layout (css).
I consider it only slightly less offensive than <blink> as being useless. I also find it incorrect from attempting to have the markup dictating the layout. Sure, this was before CSS was robust, but it is one of the reasons that I just had no problem with it going the way of the dodo.
It would be messy. If <center> element is used somewhere else then you will have to find its parent classes. If they don't have proper parent div classes or <center> used in many places then you will have to individually target <center> elements like div > div > div > center and another one div > div > div > div > center.
You can override center too but why would you try that instead of creating a class for a div and addressing it properly. Classless elements cause long-term issues when modified with CSS.
For vertical centering, I never understood why you couldn't just give the container a fixed size and then, on the element, set margin-top and margin-bottom to auto. It works like this for horizontal centering.
My guess is that the specification treats vertical margins in a weird way. For example, margin-top:50% means 50% of the width, not of the height. I don't know why they decided to do it like that.
Can somebody explain to me why centering a div is such a challenge to have an easy implementation for?
It seems like Browser makers, Languages creators, and developers all have a pretty strong incentive to solve such a fundamental, basic problem in web development to make the design experience easier.
Why haven’t these groups come together to make an easy solution to this? What am I missing?
They *have* come together. That's why we actually have several options built for purpose and native to CSS these days, as opposed to relying on hacks and Javascript (which was surprisingly common even as recently as 5-6 years ago. The name of this article is more or less a joke, because most people with any experience will recognize that since Flexbox and Grid have become widely available, the answer is generally "use Flexbox or Grid". Knowing the nuances of these layout algorithms is table stakes for building UIs that don't feel "janky".
The problem with designing and implementing application layouts for the Web is that it's such a dynamic medium. You can't ever rely on a particular viewport size, and you generally can't rely on content size either.
Like, just as a simple example: build me a page which has a white background and three blue boxes in the middle which contain white text saying "foo", "bar", and "Supercalifragilisticexpialidocious".
How big should these boxes be? Should they even be the same size? How should text wrap within them? How should the text itself be aligned within the boxes? When you say the "middle", do you mean centered vertically or horizontally? Are they laid out in a row or in a column? What's the expected behavior when the viewport size is too small to accommodate them? Should the boxes themselves wrap in some special way? Should they resize themselves?
This isn't even really splitting hairs or anything, it's just sort of the mindset you get into when you start working within a domain that's governed by *constraints* rather than specific sizes.
Because what people mean by “centring” varies depending on context. You are actually talking about a group of several different behaviours and you don’t always want the same one. So you need multiple behaviours, and you need multiple ways of specifying those behaviours.
The most obvious example: horizontal and vertical centring are different because our writing system has a specific direction it flows in. You can’t just do the same thing for horizontal as for vertical because text doesn’t work that way.
This problem was solved literally decades ago with <center>. I appreciate why that HTML tag was deprecated but I also completely sympathise with why people scoff at the complexity of doing the same thing with CSS.
this blog post and subsequent discussion is literally just about horizontal centring!
I’ve been building websites longer than most (since 1994 in fact) and centring in CSS is definitely not as easy as it was with the <center> HTML. Not even close. If it were, there wouldn’t be this entire discussion to begin with.
I swear the amount of times topics like this come up and yet a small subset of developers, likely Stockholm syndromed into believing things are great, is ridiculous. The sooner people like yourself pull your head out of your arse and realise that the current status quo is unacceptable, the sooner web standards finally mature into something that doesn’t have more footguns than the worst of C and worst of Perl combined.
I agree. However <center> (or rather a hypothetical CSS equivalent) would still more than good enough for a considerable number of common use cases.
It's perfectly reasonable to ask a particularly language to make the easy things easy while still being powerful enough to make the hard things achievable. When it comes to CSS, or even web development in general, that isn't often the case. This discussion demonstrates that point too.
If it was a markup language, it would be much easier to use. Instead, it's a rules based system that works upon a markup language (and I'm not trying to be pedantic, it would be easier if it was part of the markup, but also require much more work to change the look of content by needing to change the markup - which might be generated dynamically and difficult to change without lots of programming effort).
It's all the more ridiculous because browsers can in fact center divs (block
boxes), in an incredibly straightforward way, using the <center> tag. And all
browsers implement this using a vendor-specific CSS rule (e.g. text-align:
-moz-center). They must be able to do this, because real websites use it, and
there is no other way to implement it using CSS.
And yet it's not standardized[1], so if you implement it in a new layout engine
you're left with reverse-engineering what other browsers do (a classic).
Notice how it's not a stylesheet, not a presentational hint, just prose. It's
because standard CSS can't do what browsers could since before its invention.
I mean exactly that it's not obsolete, because there is no standard CSS
property that could have obsoleted it. (Even though every browser does have one;
it's just vendor-prefixed.)
Ok, one semantic div less. But why can't I do this?
<div style="block-align: center /* would act like text-align: -moz-center */">
<div style="width: 10ch; block-align: right /* would act like text-align: -moz-right */">
hello
</div>
</div>
If my browser understands the first variation, then getting it to understand
the last one is in all likelihood not much effort; it's just substituting
properties.
The only reason it doesn't, is that the property does not exist in the standard.
So you have to use non-semantic centers instead of semantic divs, but that's
frowned upon, so you have to use grid layout even though you aren't making
a grid, just centered flow.
I just wish it were standardized, considering it's already implemented in all
browsers (in the forms text-align: -moz/-webkit-center).
None of that has anything to do with the fact that the HTML standard--written and edited by all the major browser vendors--has declared the <center> element as obsolete.
Sorry, but if this is the only thing you were pointing out, then I don't see how
your comment is related to mine?
<center> was marked obsolete because it's not semantic HTML, and centering stuff
is the task of CSS, not HTML. It was not marked obsolete because a block-level
text-align-like property would be useless.
My comment was lamenting the fact that standard CSS does not provide a
compatible alternative, despite widespread implementation of the feature. Saying
<center> is obsolete does not resolve this issue.
As someone who jumps on-and-off of webdev it can be a little frustrating catching up on latest javascript frameworks and, at lower level, modern Javascript and CSS.
As far as “subscribe to the newsletter” prompts go, that wasn’t as annoying as what I usually see. Still, would prefer it to have been on the right side so it didn’t obscure content.
I love his posts. So much information and insight. But why the hell does a guy obviously so good at UX stuff fuck around with my scrolling? I cannot stand it.
This guy's page never pops up in google search... My only problem is, that sidekick on the page is quite irritating. Can't we have some kind of furry fantasy creature instead?
These comments are largely talking about how centring is hard from an implementation standpoint and can mean different things. Sure, but the whole "CSS centre meme" is just "How do I put one element dead centre within another". I can't point to data to back this up, but I'm pretty sure that's what 99% of people mean when they ask "Why is centring so hard?"
For those of us who grew up with GeoCities and AngelFire etc, the reason not being able to easily do this felt so ridiculous and why it was hard to buy that it's hard to implement is that we were able to do this no problem with HTML, and still can!
That puts the text "hello" squarely in the middle of a 600x600 yellow box [0]. This worked in the 90s. I know nothing about the implementation of CSS, but surely you can understand the annoyance of been handed a new tool, told it's better, yet wasn't able to do such a standard thing lots of people were already doing. It's a big reason I didn't listen to the dogma and kept up with table layouts into the late 2000s.PS, maybe it's Stockholm Syndrome, but I love CSS FWIW.
[0] You actually don't even need the `valign` which surprised me, at least not in chromium... I'm not sure if that is a recent thing because I haven't actually written code like that in around 15 years.