To a first approximation, this doesn't matter anymore. "Valid HTML" used to be a big deal because when you left the HTML spec you were inviting the various browsers to interpret your non-standard HTML in differing ways, sometimes quite catastrophically so for the styling or how the Javascript would interact with the page.
This is no longer anywhere near as important as it used to be because HTML5 defines a method for turning more-or-less any sequence of bytes into the same DOM tree: https://dev.w3.org/html5/spec-LC/parsing.html And that's only the beginning of the process. I can't seem to find a good link to the whole 8.2 section of the HTML5 spec but the whole process is freaking huge. But it's defined now.
I hedge on the "more-or-less" because I'm sure there are still bugs in various parsers and perhaps there are pathological sequences that wouldn't be handled by this process, but such sequences would be very, very distant from being HTML at all. But one difference with HTML5 is that the parsers would be considered buggy; in previous versions it could be debatable what the parser should do. HTML5 should fully specify that. If it doesn't that is now a bug in the spec. I would hope it has been banged on enough at this point that any possible remaining corner cases must be pretty small by now.
It is in my considered opinion perfectly sensible to define "HTML" as "what comes out of the HTML5 parsing process" and not really be all that worried about whether this tag does or does not need to be closed before this set of tags but not this other set of tags. It is no longer such an invitation to the browsers to render things completely differently. What was once an academic concern and a user-experience concern is now largely an academic concern.
In fact, if you're handling HTML5 correctly, which is to say, using a real, conformant parser to operate on the resulting parse tree rather than trying to handle it as a string... you can't even tell the difference between "valid" and "invalid" HTML anymore! The parser will wipe that away entirely before the HTML gets to your code. That's how important it is now.
The number is eye-catching but it's measuring how non-conforming the source code is, as opposed to whether it results in the intended DOM. This is why HTML validators largely went out fashion since it's more practical to consider source "valid" if it renders correctly in the browser. The way the browser handles non-conforming source code is part of the spec [1].
The claim in the title does not match the linked website. The 2.6% figure only includes websites that also follow some arbitrary set of "best practices" in addition to being valid HTML, some of which actually contradict best practices from the past.
As someone who valued that "Validated by w3c" button on my website back in the day, if a website today has FULLY valid HTML, I gaurantee you the admin is a massive nerd.
That's actually a good point, your markup can reveal some info about you. Like, today this kind of info is mostly unnoticed, because there are too much details in the world and too few eyeballs to notice them. But in the (not so distant) future AI agents might be able to do a lot of inference by drawing from info like "this person hand-authored their website since the 2000s and cared a lot about obscure details etc etc"
I also think it's pretty likely that if a website has fully valid HTML, that website makes little to no money. Because if it actually was a revenue-generating venture it's extremely unlikely the person or people behind it would be worrying about what tags have which attributes and which tags are self-closing and which aren't when it has zero impact on revenue.
React.js with JSX (which people use to avoid writing HTML) is basically the opposite on its tolerance for mistakes but somehow it seems pretty popular, to say the least.
JSX is a little pickier about structure, but doesn't give much care towards conformance, which is where the bulk of HTML violations occur. I expect close to 98% of React output is also invalid, which is why users are using React and not something like Qt.
Is HTML that way because of it being HTML, or have browsers just decided to brute force their way into handling the slop so that people won't blame the browser for not doing it correctly?
In my experience a lot of the "best practice" warnings are reductive.
A couple examples from this list are not requiring a type parameter on certain tags, or style tags being valid under html and body tags but not div tags. These types of things would take extra coding in a browser to check for certain singular tags not requiring things that all others do. Therefore it is simpler to leave these checks out and leave sloppy syntax checking as it does no harm at all to the rendered markup.
You may be right in some situations, but the nature of HTML having everything applied to all tags and then selectively reduced definitely helps make it a more resilient and accommodating language.
Well it's obviously a better experience for someone who's not the page author if the browser tries to render an invalid page rather than throwing up its hands and showing an error. That way you can at least try to read the page rather than not being able to and having to track down the author and try to get him to fix his site. The "strict parsing" fans tried to force correctness with XHTML, but it was a failure and didn't see much use.
The main problem with "handling the slop" used to be that different browsers would render invalid HTML differently, so poorly written sites could look very different depending on the browser you used. Modern HTML standardizes how browsers handle invalid HTML so this isn't a problem anymore. See https://html.spec.whatwg.org/multipage/parsing.html
"The start tag is required. The end tag may be omitted if the <p> element is immediately followed by an <address>, <article>, <aside>, <blockquote>, <details>, <div>, <dl>, <fieldset>, <figcaption>, <figure>, <footer>, <form>, h1, h2, h3, h4, h5, h6, <header>, <hgroup>, <hr>, <main>, <menu>, <nav>, <ol>, <pre>, <search>, <section>, <table>, <ul> or another <p> element, or if there is no more content in the parent element and the parent element is not an <a>, <audio>, <del>, <ins>, <map>, <noscript> or <video> element, or an autonomous custom element." [0][1]
Nonetheless, <p> is automatically closed by the occurrence of another block-level element, like another <p> tag or ones like <div> or any header. <li> and <tr> have similar rules. We close them because we're not savages and want to see container tags balanced in the editor, but the html5 spec doesn't strictly require it.
AIUI plenty of tags don't need to be closed. `<input>`, `<hr>`, `<img>` are the first three off the top of my head. Probably in XHTML you would've self-closed them.
50% of my errors are “<font> is deprecated”. If browsers stop supporting that I’m guessing huge swathes of the web will be affected, so I’m guessing it will never happen.
Why does it seem like Accessibility is always an after thought? It is usually so easy to do while a site is built. We just don't think about it until someone complains?
Almost nobody is testing their website with a screen reader so they don't think of it as part of the user experience. In fact, I think most developers are only vaguely aware of what web accessibility actually is for and how it affects users.
I'm curious about how many test with a screen reader even among those who do claim to care about accessibility. Most advice online around accessibility is about using alt attributes and semantic html tags or whatever, not testing your site with accessibility tools.
Making sure a site runs in more browsers than Chrome is an afterthought now.
Making sure a site can be used from a high-dropout or high-latency connection is an afterthought now.
Making sure a webapp can run effectively on anything but the development team's hottest new hardware is an afterthought now.
Making sure all the buttons and features still work is an afterthought now.
All of these things are on the same order of "easy" as accessibility, but the reality is that teams variously either don't know what they're doing or don't care what they're doing.
The norm for the last decade is to just to push out layers and layers of slop as fast as possible, optimizing on feature bulletpoints and "release cadence".
prefers-color-scheme has shipped for something like 7 years now and not even all of the major sites I frequent which already have a dark theme toggle support it yet. Obviously accessibility is more important than a dark theme (though it can be part of an accessibility story), but if it takes a minute to ship prefers-color-scheme on those existing themed sites and it's been so many years you get the idea of how little it's about when something is easy or not.
The problem with prefers-color-scheme is that it doesn't indicate an actual user preference since the default is not "no preference" (which used to exist as a possible value in at least FF before it was intentionally removed) but rather whatever the OS/browser picked. So any website that wants to choose a "default" appearance but still provide choice to users who actually care cannot use this mechanism at all.
(a) accessibility is about all disabilities, not being blind
(b) overall “significant” disability levels are more like 16% [1]
(c) the 16% is at any one time. The chances of us all experiencing disability during our lifespan are much higher
(d) accessibility is a legal requirement for many systems in many regions. Even if you’re not engaging with a region with a legal requirement yet, do you want to have to build from scratch when you do?
I would be interested to see if the metric moves meaningfully based on switching to a different set of parameters for example running from the US vs the EU and multiple attempts over multiple days to see if anything changes...
The question is how important it is to have fully valid HTML.
The HTML validator complains about a ton of useless crap IMO. At one point I simply gave up on that. Just blindly adhering to that validator makes little sense really.
What does "fully valid HTML" get you that "mostly-valid HTML" doesn't?
I'm not trying to be flippant but this just seems like a really silly thing to aim for and an even sillier thing to try to shame people about when ~98% of websites that people use just fine don't do it.
When I met my future wife, the first thing she did was to check if my personal homepage was fully validated. It was. And we are celebrating 20 raucous years this year!
To a first approximation, this doesn't matter anymore. "Valid HTML" used to be a big deal because when you left the HTML spec you were inviting the various browsers to interpret your non-standard HTML in differing ways, sometimes quite catastrophically so for the styling or how the Javascript would interact with the page.
This is no longer anywhere near as important as it used to be because HTML5 defines a method for turning more-or-less any sequence of bytes into the same DOM tree: https://dev.w3.org/html5/spec-LC/parsing.html And that's only the beginning of the process. I can't seem to find a good link to the whole 8.2 section of the HTML5 spec but the whole process is freaking huge. But it's defined now.
I hedge on the "more-or-less" because I'm sure there are still bugs in various parsers and perhaps there are pathological sequences that wouldn't be handled by this process, but such sequences would be very, very distant from being HTML at all. But one difference with HTML5 is that the parsers would be considered buggy; in previous versions it could be debatable what the parser should do. HTML5 should fully specify that. If it doesn't that is now a bug in the spec. I would hope it has been banged on enough at this point that any possible remaining corner cases must be pretty small by now.
It is in my considered opinion perfectly sensible to define "HTML" as "what comes out of the HTML5 parsing process" and not really be all that worried about whether this tag does or does not need to be closed before this set of tags but not this other set of tags. It is no longer such an invitation to the browsers to render things completely differently. What was once an academic concern and a user-experience concern is now largely an academic concern.
In fact, if you're handling HTML5 correctly, which is to say, using a real, conformant parser to operate on the resulting parse tree rather than trying to handle it as a string... you can't even tell the difference between "valid" and "invalid" HTML anymore! The parser will wipe that away entirely before the HTML gets to your code. That's how important it is now.
> "Valid HTML" used to be a big deal
To a point that there was an actual badge for it.
The number is eye-catching but it's measuring how non-conforming the source code is, as opposed to whether it results in the intended DOM. This is why HTML validators largely went out fashion since it's more practical to consider source "valid" if it renders correctly in the browser. The way the browser handles non-conforming source code is part of the spec [1].
[1] https://html.spec.whatwg.org/multipage/parsing.html#parse-er...
The claim in the title does not match the linked website. The 2.6% figure only includes websites that also follow some arbitrary set of "best practices" in addition to being valid HTML, some of which actually contradict best practices from the past.
As someone who valued that "Validated by w3c" button on my website back in the day, if a website today has FULLY valid HTML, I gaurantee you the admin is a massive nerd.
That's actually a good point, your markup can reveal some info about you. Like, today this kind of info is mostly unnoticed, because there are too much details in the world and too few eyeballs to notice them. But in the (not so distant) future AI agents might be able to do a lot of inference by drawing from info like "this person hand-authored their website since the 2000s and cared a lot about obscure details etc etc"
I also think it's pretty likely that if a website has fully valid HTML, that website makes little to no money. Because if it actually was a revenue-generating venture it's extremely unlikely the person or people behind it would be worrying about what tags have which attributes and which tags are self-closing and which aren't when it has zero impact on revenue.
Oh yeah, it's a labor of love. But it might be the corporate website of a business that makes a lot of money, not through the website though.
This is exactly why HTML is so ubiquitous, because it is so tolerant of mistakes in formatting, syntax, and just about anything else.
Most browsers will even read and process most of the things listed on this page because even if the spec doesnt say so, it just makes sense to anyway.
React.js with JSX (which people use to avoid writing HTML) is basically the opposite on its tolerance for mistakes but somehow it seems pretty popular, to say the least.
JSX is a little pickier about structure, but doesn't give much care towards conformance, which is where the bulk of HTML violations occur. I expect close to 98% of React output is also invalid, which is why users are using React and not something like Qt.
Is HTML that way because of it being HTML, or have browsers just decided to brute force their way into handling the slop so that people won't blame the browser for not doing it correctly?
In my experience a lot of the "best practice" warnings are reductive.
A couple examples from this list are not requiring a type parameter on certain tags, or style tags being valid under html and body tags but not div tags. These types of things would take extra coding in a browser to check for certain singular tags not requiring things that all others do. Therefore it is simpler to leave these checks out and leave sloppy syntax checking as it does no harm at all to the rendered markup.
You may be right in some situations, but the nature of HTML having everything applied to all tags and then selectively reduced definitely helps make it a more resilient and accommodating language.
Totally agreed.
Well it's obviously a better experience for someone who's not the page author if the browser tries to render an invalid page rather than throwing up its hands and showing an error. That way you can at least try to read the page rather than not being able to and having to track down the author and try to get him to fix his site. The "strict parsing" fans tried to force correctness with XHTML, but it was a failure and didn't see much use.
The main problem with "handling the slop" used to be that different browsers would render invalid HTML differently, so poorly written sites could look very different depending on the browser you used. Modern HTML standardizes how browsers handle invalid HTML so this isn't a problem anymore. See https://html.spec.whatwg.org/multipage/parsing.html
> An unclosed tag occurs when an opening HTML tag like <div>, <p>, or <span> is missing its corresponding closing tag.
Since when does <p> requires a closing tag? Note that TFA lists this among "spec violations" and not merely "best practices".
[delayed]
p stands for paragraph. It's not a line break
The spec on <p>:
"The start tag is required. The end tag may be omitted if the <p> element is immediately followed by an <address>, <article>, <aside>, <blockquote>, <details>, <div>, <dl>, <fieldset>, <figcaption>, <figure>, <footer>, <form>, h1, h2, h3, h4, h5, h6, <header>, <hgroup>, <hr>, <main>, <menu>, <nav>, <ol>, <pre>, <search>, <section>, <table>, <ul> or another <p> element, or if there is no more content in the parent element and the parent element is not an <a>, <audio>, <del>, <ins>, <map>, <noscript> or <video> element, or an autonomous custom element." [0][1]
[0] https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/...
[1] https://html.spec.whatwg.org/#the-p-element
Nonetheless, <p> is automatically closed by the occurrence of another block-level element, like another <p> tag or ones like <div> or any header. <li> and <tr> have similar rules. We close them because we're not savages and want to see container tags balanced in the editor, but the html5 spec doesn't strictly require it.
I always assumed all tags should be closed.
AIUI plenty of tags don't need to be closed. `<input>`, `<hr>`, `<img>` are the first three off the top of my head. Probably in XHTML you would've self-closed them.
HN is not one of them https://validator.w3.org/nu/?doc=https%3A%2F%2Fnews.ycombina...
Seems the majority of it is just using obsolete tags, and suggesting that CSS be used instead lol
Well, if it ain't broke...
50% of my errors are “<font> is deprecated”. If browsers stop supporting that I’m guessing huge swathes of the web will be affected, so I’m guessing it will never happen.
I wonder how many of them only work because of things like https://github.com/WebKit/WebKit/blob/main/Source/WebCore/pa...
Which overlaps with, but isn't the same as, JS quirks mode.
I would def argue this is a feature of the net ( that people from all walks of browsers can see mostly the same thing ) and not a bug.
Almost nobody is testing their website with a screen reader so they don't think of it as part of the user experience. In fact, I think most developers are only vaguely aware of what web accessibility actually is for and how it affects users.
I'm curious about how many test with a screen reader even among those who do claim to care about accessibility. Most advice online around accessibility is about using alt attributes and semantic html tags or whatever, not testing your site with accessibility tools.
Ironically the site does not support `:prefers-color-scheme`.
prefers-color-scheme is not an accessibility feature. The idea that dark mode matters for accessibility is a meme.
Dark mode is bad for people with astigmatic halation.
Light mode, I've read, is worse for people with cataracts.
Visual accessibility doesn't just apply to the totally blind. There's a reason things like WCAG exist.
Making sure a site runs in more browsers than Chrome is an afterthought now.
Making sure a site can be used from a high-dropout or high-latency connection is an afterthought now.
Making sure a webapp can run effectively on anything but the development team's hottest new hardware is an afterthought now.
Making sure all the buttons and features still work is an afterthought now.
All of these things are on the same order of "easy" as accessibility, but the reality is that teams variously either don't know what they're doing or don't care what they're doing.
The norm for the last decade is to just to push out layers and layers of slop as fast as possible, optimizing on feature bulletpoints and "release cadence".
prefers-color-scheme has shipped for something like 7 years now and not even all of the major sites I frequent which already have a dark theme toggle support it yet. Obviously accessibility is more important than a dark theme (though it can be part of an accessibility story), but if it takes a minute to ship prefers-color-scheme on those existing themed sites and it's been so many years you get the idea of how little it's about when something is easy or not.
The problem with prefers-color-scheme is that it doesn't indicate an actual user preference since the default is not "no preference" (which used to exist as a possible value in at least FF before it was intentionally removed) but rather whatever the OS/browser picked. So any website that wants to choose a "default" appearance but still provide choice to users who actually care cannot use this mechanism at all.
Because 0.5% of the world population is blind and it costs much more than an extra 0.5% to make a website accessible?
Everytime this question is asked I wonder if people are genuinely not understanding that. It's just not financially worth it.
I'm not defending this position, but it's just glaringly obvious.
(a) accessibility is about all disabilities, not being blind
(b) overall “significant” disability levels are more like 16% [1]
(c) the 16% is at any one time. The chances of us all experiencing disability during our lifespan are much higher
(d) accessibility is a legal requirement for many systems in many regions. Even if you’re not engaging with a region with a legal requirement yet, do you want to have to build from scratch when you do?
[1] https://www.who.int/health-topics/disability
Accessibility is about way more than just people who are blind.
I would be interested to see if the metric moves meaningfully based on switching to a different set of parameters for example running from the US vs the EU and multiple attempts over multiple days to see if anything changes...
The question is how important it is to have fully valid HTML.
The HTML validator complains about a ton of useless crap IMO. At one point I simply gave up on that. Just blindly adhering to that validator makes little sense really.
Honestly more than I thought
And not a single person cares. What does fully valid even mean anymore?
[flagged]
Legit who cares? It's not stopping people from visiting them
At this point, with all the tools available, if your site does not have fully valid HTML, it is because you don't care. LLMs are your friends, dev.
What does "fully valid HTML" get you that "mostly-valid HTML" doesn't?
I'm not trying to be flippant but this just seems like a really silly thing to aim for and an even sillier thing to try to shame people about when ~98% of websites that people use just fine don't do it.
When I met my future wife, the first thing she did was to check if my personal homepage was fully validated. It was. And we are celebrating 20 raucous years this year!
Honestly? I have no idea. But, hey, if it works, it works.
[dead]