Note that I use paper, because I am an old person. They're very different problems.

In print, you have to deal with page breaks and positioning of figures, but you know everyone will see the same thing. Doing this right requires human attention, but you can produce beautiful layouts, even with multiple columns and oddly-shaped text boxes.

On the web, everyone sees a single "page," but those pages are all different. Even simple two-column justified text is mostly garbage on the web, and forget trying to float figures close to their references. Web designers have mostly given up, resorting to either huge margins, or an ugly wide wall of giant-font text. Print-quality layout and typesetting on the web is very far away.

Completely true, but for fairness the reason is that the web was not created as a tool for text layout. The goal was exactly the opposite, to make the text independent of layout and displayed differently in the different browsers. Doing good text layout on the web is practically an intractable problem. On the other hand, we already have software capable of doing this, such as PDF.

I don't know if Tim Berners-Lee thought about layout. SGML was text with semantic information, and HTML was text with cross-references. Back when displays were tiny low-resolution things little better than terminals, and most web pages were just text with a few images, layout didn't matter. It was all ugly. Now that displays are larger than books, and pages are heavy enough to contain the entire Gutenberg Bible, people are starting to care.

Maybe in 5-10 years, with enough server-side computing power and bandwidth, sites will ask for the client's screen size and desired font size, then serve an appropriate PDF.