I'm going to take the stance of: "I don't see the usefulness of it, but that doesn't mean it can't be useful." Maybe it's working within the limitations of a particular language that doing something like this is the best way. Maybe I'm completely missing some clear and obvious sign.

> The principle of lowest friction is that this templating system must, at the very minimum, match, if not exceed the ease of HTML.

Realizing also you mentioned as it grows larger, abstraction occurs, this is where I don't see the value. For anyone to use this templating system, they need to know HTML and then they need to know this templating system (which amounts to learning a programming language). Mistakes in the templating system will result in HTML errors. Maybe I've just had too many bad experience with systems like this where fixing the HTML would have been easy, except I had to go through a templating system that abstracted away critical parts.

I'll assume I'm just missing something, or I've had this problem your solving before, and solved it in some other way. Like I said, I don't see how this method benefits anything (and I've done my fair share of work on the web), but I'll chalk that up to a failing on my part to grasp some concept. =)