[Mikhail_Edoshin]

Semantic in machine processing is actually very simple: if a machine has an instruction to process an element and we know what it does, then the element is semantic.

So, for example, <b> and <i> have perfect semantic, while <article> not so much. What does the browser do with an <article>? Or maybe it is there for an indexing engine? I myself have no idea (nor that I investigated that, I admit).

But all that was misunderstood, very much like XML itself.

[zzo38computer]

The <article> command in HTML can be useful, even if most implementations do not do much with it. For example, a browser could offer the possibility to print or display only the contents of a single <article> block, or to display marks in the scrollbar for which positions in the scrollbar correspond to the contents of the <article> block. It would also be true of <time>; although many implementations do not do much with it, they could do stuff with it. And, also of <h1>, <h2>, etc; although browsers have built-in styles for them, allowing the end user to customize them is helpful, and so is the possibility of using them to automatically display the table of contents in a separate menu. None of these behaviours should need to be standardized; they can be by the implementation and by the end user configuration etc; only the meaning of the commands will be standardized, not their behaviour.

[Mikhail_Edoshin]

"Meaning" has a rather vague meaning, but behavior is exact. If I know the behavior, it becomes a tool I can employ. If I only know supposed behavior, I cannot really use that. E.g. why we have so much SEO slop and so little "semantic" HTML? Because the behavior of search engines is real and thus usable, even when it is not documented much.