> HTML semantics and CSS quality are generally in a disastrous state
I believe many people have tried hard to do the right thing. But the truth of the matter is that HTML as a markup vocabulary is stuck at a 80's core language with a 90's vision for casual academic publishing mixed in, and then since mid/late 90's, CSS (and a gazillion of divs and spans) has been kludged on top to make up for HTML's lack of evolution and inappropriateness to describe digital content in this millenium. Yes we can kindof make it work, but it cannot be said that today's web stack is even remotely usable by actual content creators like even Flash used to be. Instead, we're churning through complex JavaScript frameworks, new languages, and entire new runtimes to exercise browsers into a workable platform for web apps with a laughable power budget.
What's left of grand visions for "semantic" HTML and structure/presentation separation, invented to lend credit to the HTML/CSS/JS trifecta after the fact more than anything else IMO, is maybe accessibility to some degree (which is at least something) and a hope for well-meant, bona fide SEO; and even the latter isn't helping anyone, and hasn't for a long time.
The fundamental flaw of the proposed structure/presentation dichotomy is that a piece of digital content should be captured close to the "intent" of its author, which in the majority of cases is very different from an 80's idea of an academic paper with section, subsections, and paragraphs.
This despite HTML being based on SGML which had always all means necessary to evolve markup in a highly structured fashion, and even encourages using your own vocabulary to map into an output/presentation vocabulary. For example, to model a sub-vocabulary for capturing a discussion thread like the one we're having here, which seems quite a natural use case for HTML - but nope, we must use hierarchical text and then pile a shit-tonne of CSS over it. Is that really "semantic", or rather a case of eternal September and embarrassing self-delusion?
This is a great comment. I'm pretty sure every non-trivial website would be better specified (as opposed to HTML) in almost any measurable way (LOC, clarity, etc) in say raw opengl.
Yes, creating a UI in an immediate mode tech would be pretty hard - still easier than HTML/CSS.
I'd love to see someone prove me right: Take any non-trivial site like twitter, facebook.com, whatever and reproduce it with opengl/d3d or whatever. I bet it will be less LOC (incl all the libs the source website includes), more clear, easier to extend, faster to draw, et cetera.
I’d love to see someone try and epicly fail to even render legible text. I have a feeling this thread is seriously underappreciating how many things browsers get right for them.
To be clear, I think the authors of the browsers have in general done a heroic job. I'm not talking about the browser impl, I'm impugning HTML and CSS as UI specification mechanisms.
I believe many people have tried hard to do the right thing. But the truth of the matter is that HTML as a markup vocabulary is stuck at a 80's core language with a 90's vision for casual academic publishing mixed in, and then since mid/late 90's, CSS (and a gazillion of divs and spans) has been kludged on top to make up for HTML's lack of evolution and inappropriateness to describe digital content in this millenium. Yes we can kindof make it work, but it cannot be said that today's web stack is even remotely usable by actual content creators like even Flash used to be. Instead, we're churning through complex JavaScript frameworks, new languages, and entire new runtimes to exercise browsers into a workable platform for web apps with a laughable power budget.
What's left of grand visions for "semantic" HTML and structure/presentation separation, invented to lend credit to the HTML/CSS/JS trifecta after the fact more than anything else IMO, is maybe accessibility to some degree (which is at least something) and a hope for well-meant, bona fide SEO; and even the latter isn't helping anyone, and hasn't for a long time.
The fundamental flaw of the proposed structure/presentation dichotomy is that a piece of digital content should be captured close to the "intent" of its author, which in the majority of cases is very different from an 80's idea of an academic paper with section, subsections, and paragraphs.
This despite HTML being based on SGML which had always all means necessary to evolve markup in a highly structured fashion, and even encourages using your own vocabulary to map into an output/presentation vocabulary. For example, to model a sub-vocabulary for capturing a discussion thread like the one we're having here, which seems quite a natural use case for HTML - but nope, we must use hierarchical text and then pile a shit-tonne of CSS over it. Is that really "semantic", or rather a case of eternal September and embarrassing self-delusion?