For one thing, I just genuinely think WebComponents are a badly designed API that is weird and hard to use (how many people are using WebComponents without at least Lit, if not something much bigger?), and React is a relatively well-designed library that isn't really that bloated. There's not really much of a point in trying to argue since this is inherently subjective and people with different values are going to irreconcilably disagree. But, if you don't respect that some people hold this position, we're not going to make any progress towards a consensus.
On the note of <dialog>, I recently tried to use <dialog> in a (React) application, and it did work pretty well, but I also found that in Firefox it is only practically possible to do a fade-in animation, not a fade-out one. That isn't really a critical issue for me, it is just an animation after all, but I find it unfortunate. I also find <dialog> to be a weirdly shaped API too: I don't really hate it, but I don't love it either. It feels awkward.
I find this implicit view that developers that, for example, prefer React over WebComponents are making a suboptimal choice to be rather condescending and not really in the spirit of trying to see things from the other side. Wouldn't you want to focus on the strongest arguments and not the weakest ones? Maybe you've literally never heard anyone complain about WebComponents or Shadow DOM, but if so, I find that surprising. Certainly here on HN, I've seen a fair bit of WebComponents hate.
I do, FWIW, realize that I've particularly focused on WebComponents, which this article doesn't actually name directly. But, I assume we're not talking about ditching React to implement our own component framework on top of the traditional DOM APIs, because that's what React already does...
> React is a relatively well-designed library that isn't really that bloated.
I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior. But it leaves the question of why react is still so popular. I think the biggest reason is all the non-technical aspects of react:
- They have excellent documentation. And have, from day 1.
- They produced videos, sample projects, and all sorts of "getting started" documentation.
- They ran react conferences, teaching everyone who would listen about "1 way data flows" and pretending like they invented FP.
The amount of hype around it made it really feel like the next big thing. Between the very well funded react team and the outside developer community, there was real momentum. People learned it in droves. Taught students. Built websites with it. When react's poor design choices caused issues (and there were a lot of issues), then you were blamed for holding it wrong. (Component classes, state, CSS, hooks, webpack and babel taking ages, big bundle sizes, slow re-renders, and so on.)
By the time the next generation of JS frameworks broke onto the scene, there was a collective moan from the community. "Oh no, not again - we just relearned how to make websites." React was the wave, and in its wake we all had Framework fatigue.
Software doesn't get popular without a lot of work by dedicated people. I really admire the work standards bodies do. But they rarely bother to take the time to produce documentation, videos, tutorials, starter projects, blogs and podcasts and all the rest of that work.
> I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior.
I don't really think that React is necessarily the best possible library, but there is certainly more to a UI library than simply compiling out the need for diffing. For one thing, I find JSX to be a relatively unobtrusive addition to the language. It is implemented by a variety of things and is a pretty simple transform as far as things go; basically pure sugar, it's possible to avoid it if you want. It could be better designed than it is, but I think it's at least sufficiently general - it is feasible to say, use an alternate library with JSX, like Preact.
Svelte on the other hand by its nature just simply requires a more complicated compiler step to work and deliver on its promises. That's a trade-off. Whether you think it's worth it is up to you, but presenting it as an objectively technically superior design is disingenuous. Superior how? Runtime costs? Sure, but it does beg the question how much. Will a well-optimized Svelte app feel much better than a well-optimized React app?
> The amount of hype around it made it really feel like the next big thing.
We are a really long way away from the hype of React. I would argue that hype stopped carrying React a long time ago. Momentum? Sure, but if something was truly dramatically better, then I think it would have a fair shake at beating out React.
The problem is that the web isn't microbenchmarks and developer experience does matter to some degree, so the trade-offs that seem so good on paper don't always pan out.
I'm not a hater of Svelte. On paper, it is a very elegant idea. However, when I actually tried to use it, I genuinely came out feeling that it was simply not for me, and the benefits it has are not enticing enough for me to continue experimenting. My React apps are not particularly slow or bad at handling huge amounts of data. (One of my apps has no problem handling a file grid of over 1 million items, and in fact, I run into issues with browsers not being able to handle a large enough scroll area long before performance of my React code is a problem. That is good enough for me.)
> I just genuinely think WebComponents are a badly designed API that is weird and hard to use
This was exactly what sprang to mind as soon as I saw the title. I’ve been writing front-end code for over 25 years now, and exactly two things in that entire time have made me miserable enough to think about stopping: Internet Explorer 6 and web components. Every time I try to just “use the platform” it makes me miserable and I end up demotivated and stop working on whatever side project I chose to try again with. And I otherwise like the web platform! I’ve been building with it since there was nothing but the web platform. But web components kill my enthusiasm for it stone dead.
Even now in the age of agentic development, AI trips over all the same footguns in web components that humans do. It just seems like everybody involved has been adding to the standards with “yes, and…” without ever thinking about how it will be used by web developers in practice.
I have stubbornly avoided React. When native WebComponents became a thing, I figured I'd try them out on a small internal site I maintain. It was very slow. In an effort to reduce a bit of duplicate code, my site when from loading instantly to having a noticeable load time. Having to employ tricks to try and optimize and speed up the most basic implementation of a built-in feature seemed like the wrong move, so I ditched them all together and reverted back to my old structure.
As far as I know there are still some things "the platform" actually won't deliver; for instance I'm fairly certain there's no searchable combobox that's fully accessible. One ~must leave the platform to give users the experience the expect.
That means "the platform" itself is actually teaching people to work "off platform" and arguably will always have gaps as functionality and expectations evolve.
Admittedly, the habit really ought to be 'I need to make a combobox -> does the platform have what I need? -> Research, evaluate, test -> Otherwise, make it' but I don't begrudge people for skipping the middle steps.
I actually still use Dragula. I can't quite put my finger on it, but it just feels better than Sortable. I may be imagining it, but I also don't care because, you know, it's just a JS library.
I feel like it’s a historical accident. In the past the platform couldn’t do it all, and if you wanted a dialog you had to use a library. Developers were trained to reach for libraries when they needed something like a dialog.
Then react came along and all developed learning web development after react had little to no knowledge of the platform. They were taught that touching the DOM was a bad thing to do.
That’s about it. Even now when I advocate for vanilla css I get the side eye and “tailwind and shadcn should be the default”. I get it, it’s what most people are familiar with, even if it’s worse than the platform.
Depends on the criteria, e.g. if dev speed has highest prio, a full blown component lib like Material UI would be superior to shadcn. CSS can obviously do things that tailwind can't, so it is more powerful.
Because the platform sucks. I want a rounded button with a certain radius and a certain font, that feels squishy and satisfying when you press it, especially on mobile.
The last couple of websites I've made have been largely the output of Claude with some instructions to keep to WCAG AAA accessibility standards and to optimize for loading and rendering times, and it's done a pretty decent job. If you're insistent about page weight it will avoid adding JS and React and use to browser-native elements, CSS, and vanilla JS where it can.
I think the problem is that you have to ask, and to know the language to get the result you're after. If you just ask for a pretty website you're getting 800KB of React libraries to render something, and all in AI Beige with Inter as your font choice.
It’s funny, the article proceeds to answer the question in great detail: familiarity, level of abstraction, documentation, etc.
I agree with his point that learning to do it yourself can be fun and lead to a flywheel of improvement where the next time you’re faster and the next time faster still.
In the past that was the recipe for success as a developer, nowadays with AI many feel that’s no longer the case.
It's because frankly the platform has been historically shit. Web components comes to mind as someone else here has mentioned, as it's just not a good API and needs wrapping to make it usable, same as IndexedDB which the author readily admits as a creator of such tooling. It seems like the ones who determine the platform aren't actually developers day in and day out and thus don't dogfood their own products while the rest of us do, leading us to reinvent the wheel.
Here is a great comment and discussion (scroll down) by Ian Hickson who wrote the HTML5 spec on how the platform has essentially failed on its promises causing everyone to write everything in JS or even other paradigms like Flutter and other WASM or canvas based frameworks that eschew the web entirely: https://news.ycombinator.com/item?id=34612696
Is it just me or does it feel like Anthropic and OpenAI (and probably others) are pumping harness engineering and the like to try and convince engineering organizations, or even entire companies, to "use the platform", and not meddle with software anymore?
Because the platform is too hard. The average person does not want to spend time figuring out how to get stuff to work. All the browser documentation is 1000 times nerdier than the average person wants. They want all this low level stuff abstracted away from them.
Even if you want to do it the nerdy way and bust out a text editor and start writing some HTML tags it all looks ugly out of the box. Browsers pushed everything for naming websites onto 3rd party devs, so they shouldn't be surprised when they want to use 3rd party dev's stuff over the bad and confusing 1st party ones. The browser developers are in an ivory tower building a product that doesn't really care about web developers and their needs.
since the author brought up the native <dialog>, developers chose custom implementations of dialogs because they want better accessibility, better focus management, better mobile and touch screen reader support, and more flexibility. adobe's implementation is far more robust and flexible compared to the native dialog, and it's hard to justify "use the platform" when it results in a strictly worse end result.
Yeah, that was kind of ignored in the article and it's definitely an issue for some platform features. Date pickers are another one where it's easy to outgrow the native implementation.
I was excited about the native date picker and tried to use it once, only to discover that you couldn't (can't still, I assume) disable dates other than by setting min and max. Every native browser widget I've tried to use doesn't implement the features my clients expect, so I don't use them. People want a better web platform and the only way to get it is JavaScript.
But also custom date pickers are the most broken thing you can find everywhere. Browse with a combination of browser/phone the developer didn't test, and you can't pick even the most basic single date. The most blatant are the fancy range pickers, instead of two date picker boxes: I have been pushed to a desktop browser because the mobile renders half of it off-screen, or don't hold open after tapping the start date, or works by dragging but breaks on single taps...
For one thing, I just genuinely think WebComponents are a badly designed API that is weird and hard to use (how many people are using WebComponents without at least Lit, if not something much bigger?), and React is a relatively well-designed library that isn't really that bloated. There's not really much of a point in trying to argue since this is inherently subjective and people with different values are going to irreconcilably disagree. But, if you don't respect that some people hold this position, we're not going to make any progress towards a consensus.
On the note of <dialog>, I recently tried to use <dialog> in a (React) application, and it did work pretty well, but I also found that in Firefox it is only practically possible to do a fade-in animation, not a fade-out one. That isn't really a critical issue for me, it is just an animation after all, but I find it unfortunate. I also find <dialog> to be a weirdly shaped API too: I don't really hate it, but I don't love it either. It feels awkward.
I find this implicit view that developers that, for example, prefer React over WebComponents are making a suboptimal choice to be rather condescending and not really in the spirit of trying to see things from the other side. Wouldn't you want to focus on the strongest arguments and not the weakest ones? Maybe you've literally never heard anyone complain about WebComponents or Shadow DOM, but if so, I find that surprising. Certainly here on HN, I've seen a fair bit of WebComponents hate.
I do, FWIW, realize that I've particularly focused on WebComponents, which this article doesn't actually name directly. But, I assume we're not talking about ditching React to implement our own component framework on top of the traditional DOM APIs, because that's what React already does...
> React is a relatively well-designed library that isn't really that bloated.
I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior. But it leaves the question of why react is still so popular. I think the biggest reason is all the non-technical aspects of react:
- They have excellent documentation. And have, from day 1.
- They produced videos, sample projects, and all sorts of "getting started" documentation.
- They ran react conferences, teaching everyone who would listen about "1 way data flows" and pretending like they invented FP.
The amount of hype around it made it really feel like the next big thing. Between the very well funded react team and the outside developer community, there was real momentum. People learned it in droves. Taught students. Built websites with it. When react's poor design choices caused issues (and there were a lot of issues), then you were blamed for holding it wrong. (Component classes, state, CSS, hooks, webpack and babel taking ages, big bundle sizes, slow re-renders, and so on.)
By the time the next generation of JS frameworks broke onto the scene, there was a collective moan from the community. "Oh no, not again - we just relearned how to make websites." React was the wave, and in its wake we all had Framework fatigue.
Software doesn't get popular without a lot of work by dedicated people. I really admire the work standards bodies do. But they rarely bother to take the time to produce documentation, videos, tutorials, starter projects, blogs and podcasts and all the rest of that work.
> I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior.
I don't really think that React is necessarily the best possible library, but there is certainly more to a UI library than simply compiling out the need for diffing. For one thing, I find JSX to be a relatively unobtrusive addition to the language. It is implemented by a variety of things and is a pretty simple transform as far as things go; basically pure sugar, it's possible to avoid it if you want. It could be better designed than it is, but I think it's at least sufficiently general - it is feasible to say, use an alternate library with JSX, like Preact.
Svelte on the other hand by its nature just simply requires a more complicated compiler step to work and deliver on its promises. That's a trade-off. Whether you think it's worth it is up to you, but presenting it as an objectively technically superior design is disingenuous. Superior how? Runtime costs? Sure, but it does beg the question how much. Will a well-optimized Svelte app feel much better than a well-optimized React app?
> The amount of hype around it made it really feel like the next big thing.
We are a really long way away from the hype of React. I would argue that hype stopped carrying React a long time ago. Momentum? Sure, but if something was truly dramatically better, then I think it would have a fair shake at beating out React.
The problem is that the web isn't microbenchmarks and developer experience does matter to some degree, so the trade-offs that seem so good on paper don't always pan out.
I'm not a hater of Svelte. On paper, it is a very elegant idea. However, when I actually tried to use it, I genuinely came out feeling that it was simply not for me, and the benefits it has are not enticing enough for me to continue experimenting. My React apps are not particularly slow or bad at handling huge amounts of data. (One of my apps has no problem handling a file grid of over 1 million items, and in fact, I run into issues with browsers not being able to handle a large enough scroll area long before performance of my React code is a problem. That is good enough for me.)
What do you not like about WebComponents?
Say we want to make an icon that when clicked shows how often it was clicked. The webcomponent code seens quite sane to me:
Try it here:https://plnkr.co/edit/0XUOLyM52xfFiIBu?open=index.html&previ...
> I just genuinely think WebComponents are a badly designed API that is weird and hard to use
This was exactly what sprang to mind as soon as I saw the title. I’ve been writing front-end code for over 25 years now, and exactly two things in that entire time have made me miserable enough to think about stopping: Internet Explorer 6 and web components. Every time I try to just “use the platform” it makes me miserable and I end up demotivated and stop working on whatever side project I chose to try again with. And I otherwise like the web platform! I’ve been building with it since there was nothing but the web platform. But web components kill my enthusiasm for it stone dead.
Even now in the age of agentic development, AI trips over all the same footguns in web components that humans do. It just seems like everybody involved has been adding to the standards with “yes, and…” without ever thinking about how it will be used by web developers in practice.
What about the web components APIs are bad, and how could they be done differently given the reality of the existing platform?
Only if you mean React before they went crazy with hooks and "use whatever" magic strings.
I put up with it, because plenty SaaS products favour Next.js and React as their only extension SDK.
I would never pick it up freely, all side projects with Web are VanilaJS, coupled with what is available on Java, .NET and PHP.
I have stubbornly avoided React. When native WebComponents became a thing, I figured I'd try them out on a small internal site I maintain. It was very slow. In an effort to reduce a bit of duplicate code, my site when from loading instantly to having a noticeable load time. Having to employ tricks to try and optimize and speed up the most basic implementation of a built-in feature seemed like the wrong move, so I ditched them all together and reverted back to my old structure.
As far as I know there are still some things "the platform" actually won't deliver; for instance I'm fairly certain there's no searchable combobox that's fully accessible. One ~must leave the platform to give users the experience the expect.
That means "the platform" itself is actually teaching people to work "off platform" and arguably will always have gaps as functionality and expectations evolve.
Admittedly, the habit really ought to be 'I need to make a combobox -> does the platform have what I need? -> Research, evaluate, test -> Otherwise, make it' but I don't begrudge people for skipping the middle steps.
An aside: This name an logo are incredible. https://bevacqua.github.io/dragula/
I actually still use Dragula. I can't quite put my finger on it, but it just feels better than Sortable. I may be imagining it, but I also don't care because, you know, it's just a JS library.
I feel like it’s a historical accident. In the past the platform couldn’t do it all, and if you wanted a dialog you had to use a library. Developers were trained to reach for libraries when they needed something like a dialog.
Then react came along and all developed learning web development after react had little to no knowledge of the platform. They were taught that touching the DOM was a bad thing to do.
That’s about it. Even now when I advocate for vanilla css I get the side eye and “tailwind and shadcn should be the default”. I get it, it’s what most people are familiar with, even if it’s worse than the platform.
How is tailwind and shadcn worse than platform? It speeds up development by orders of magnitude.
Depends on the criteria, e.g. if dev speed has highest prio, a full blown component lib like Material UI would be superior to shadcn. CSS can obviously do things that tailwind can't, so it is more powerful.
Because the platform sucks. I want a rounded button with a certain radius and a certain font, that feels squishy and satisfying when you press it, especially on mobile.
The last couple of websites I've made have been largely the output of Claude with some instructions to keep to WCAG AAA accessibility standards and to optimize for loading and rendering times, and it's done a pretty decent job. If you're insistent about page weight it will avoid adding JS and React and use to browser-native elements, CSS, and vanilla JS where it can.
I think the problem is that you have to ask, and to know the language to get the result you're after. If you just ask for a pretty website you're getting 800KB of React libraries to render something, and all in AI Beige with Inter as your font choice.
can the title be changed to "Why don't more developers use browser native features?"
It’s funny, the article proceeds to answer the question in great detail: familiarity, level of abstraction, documentation, etc.
I agree with his point that learning to do it yourself can be fun and lead to a flywheel of improvement where the next time you’re faster and the next time faster still.
In the past that was the recipe for success as a developer, nowadays with AI many feel that’s no longer the case.
Extreme apathy towards "learn to do it yourself" is exactly why "AI" is so popular right in programming right now.
It's because frankly the platform has been historically shit. Web components comes to mind as someone else here has mentioned, as it's just not a good API and needs wrapping to make it usable, same as IndexedDB which the author readily admits as a creator of such tooling. It seems like the ones who determine the platform aren't actually developers day in and day out and thus don't dogfood their own products while the rest of us do, leading us to reinvent the wheel.
Here is a great comment and discussion (scroll down) by Ian Hickson who wrote the HTML5 spec on how the platform has essentially failed on its promises causing everyone to write everything in JS or even other paradigms like Flutter and other WASM or canvas based frameworks that eschew the web entirely: https://news.ycombinator.com/item?id=34612696
Is it just me or does it feel like Anthropic and OpenAI (and probably others) are pumping harness engineering and the like to try and convince engineering organizations, or even entire companies, to "use the platform", and not meddle with software anymore?
Because the platform is too hard. The average person does not want to spend time figuring out how to get stuff to work. All the browser documentation is 1000 times nerdier than the average person wants. They want all this low level stuff abstracted away from them.
Even if you want to do it the nerdy way and bust out a text editor and start writing some HTML tags it all looks ugly out of the box. Browsers pushed everything for naming websites onto 3rd party devs, so they shouldn't be surprised when they want to use 3rd party dev's stuff over the bad and confusing 1st party ones. The browser developers are in an ivory tower building a product that doesn't really care about web developers and their needs.
since the author brought up the native <dialog>, developers chose custom implementations of dialogs because they want better accessibility, better focus management, better mobile and touch screen reader support, and more flexibility. adobe's implementation is far more robust and flexible compared to the native dialog, and it's hard to justify "use the platform" when it results in a strictly worse end result.
Yeah, that was kind of ignored in the article and it's definitely an issue for some platform features. Date pickers are another one where it's easy to outgrow the native implementation.
I was excited about the native date picker and tried to use it once, only to discover that you couldn't (can't still, I assume) disable dates other than by setting min and max. Every native browser widget I've tried to use doesn't implement the features my clients expect, so I don't use them. People want a better web platform and the only way to get it is JavaScript.
But also custom date pickers are the most broken thing you can find everywhere. Browse with a combination of browser/phone the developer didn't test, and you can't pick even the most basic single date. The most blatant are the fancy range pickers, instead of two date picker boxes: I have been pushed to a desktop browser because the mobile renders half of it off-screen, or don't hold open after tapping the start date, or works by dragging but breaks on single taps...
Yep.
“We need a date picker to schedule a visit.”
“Okay, use `<input type=date>`”
“It needs to be a date in the future.”
“No problem, add a `min="2026-10-05"` attribute.”
“And we’re only available on weekdays.”
“Okay, throw the native date picker away completely and build one yourself from scratch.”
Did nobody involved ever bother to ask what the requirements were for a date picker? Excluding dates is one of the most common requirements there is!