You do remember the hell that was the font tag, don’t you?

I think a lot of confusion from HTML being used for UIs vs hypertext documents. It’s a natural evolved mess still balancing the two.

But style should apply to the applied representation, while HTML should represent the intent of the content. The line is fuzzy, and HTML/CSS has a bad history of what falls where.

The cool thing with browsers is that we can bake in style to support multiple representations at once: monitor, mobile, printer, etc. Using the style attribute directly is mostly a side effect of only targeting the browser. Although your example of relative units is more ambiguous. Using a proper tag to represent the intent or falling back to a class where HTML lacks one would be preferable.

I think the separation makes more sense if you’ve used other document systems. SGML or XML DocBook, LaTeX, ROFF, etc.

For the last, if I use the man or mandoc module, all the man page norms are there for me and I’m just plugging in the content with the right macros.

W/o, I’m manually flipping registers all over the place.

But I’ve never hear of a roff or LaTeX UI, so…

Neither is necessarily "better", but rather they are different design choices. HTML at first wasn't intended to have very much styling at all, and originally there was some more styling mixed into the HTML itself (like <font>, bgcolor, etc.). Since now we pretty much want to make our webpages look like anything, it seemed to make sense to those drafting the standards to separate the two concerns, since HTML is mostly about page flow structure and data display, and page sources would be even harder to read than they already are if everyone was just putting their styles within HTML tags. If you're looking for a data attribute, it can be hard on the eyes to be constantly sifting through attributes for styling.

You certainly can do things to overload attributes and do what you're suggesting (in a sort of hackish way), and you wouldn't necessarily be wrong in doing so if that's your preference. Many people just don't like that.

One of the great things about separating styles out of HTML is that they can be contained in files and included separately. With poor connectivity, ideally, it's beneficial to keep the main page source to be as slim as possible so that it can still be downloaded and read even if the styling hasn't yet finished downloading. That's not nearly as relevant anymore, but it would not be surprising if keeping pages lean was a motivator in moving away from putting more stuff in HTML.

Contrary to popular belief, HTML and CSS are unrelated.

If any CSS property was also an HTML attribute, any addition to either spec would have to make sure it wouldn’t conflict. One example is the `content` HTML attribute and CSS property. They serve different purposes and have a completely different syntax.

    <meta name="description" content="Cool stuff”>

and

    .css::before {
        content: url(image.bmp)
    }

It keeps style instructions together instead of mingling them with data and behaviour attributes. There's some exceptions with some elements having attributes like width and border but it's mostly historical.