In human-machine interaction "semantic" means "you know exactly what the computer will do with that". So "<i>" or "<b>" tags have very clear semantic, while "<e

In human-machine interaction "semantic" means "you know exactly what the computer will do with that". So "<i>" or "<b>" tags have very clear semantic, while "<em>" or "<strong>" are less so and things like "<article>" are totally vague. Although this is not quite correct. "Semantic" is what makes sufficient distinctions for a given case. For example, assume I'm going to publish Ashby's "Introduction to cybernetics". It has an interesting structure: numbered chapters ("2"), smaller divisions within a chapter with a title but without a number, and yet more smaller units that are numbered like "2/17". Some of these smaller units have a title, some just a number. The point is that all this is rather unique. So I'm about to mark up the text to indicate all this. A semantic way to do that would be to invent a notation that is as unique as this book. There will be "<ashby-chp>", "<ashby-div>", "<ashby-unit>" and such. We do not specify how this is going to be rendered, but at least we faithfully describe what we have without losing anything and without adding anything irrelevant. Now we want to render it on a visual media. Here we have a different notation that describes fonts, styles, spacing and so on. These are very different distinctions from the structure of the book, but are very approriate for typesetting. We decide how we represent our "ashby" distinctions with these tools and write a transform from our notation into the visual one. Same for a screen reader. Here we have a notation that describes pitch, speed, etc. No spacing or fonts, of course. We decide how we are going to represent our distinctions with these and come up with another transform. At each step each element in our notations has a clear purpose. The "ashby" notation captures all the distinctions the author needs to make his point. The "visual" and "aural" notations are tools to express distinctions that can be made on a specific media. This is semantic. And this, by the way, is the original idea of XML (a multitude of notations) and XSLT (the notation transformer). [And the description of the transforms also uses yet another notation with yet another clear purpose :)]