I see your points, but I disagree with highlighting being provided externally. The text rendering surface ought to be able to do it -- there are plenty of libraries that do a decent job of language auto-detection, for instance, all the more so in the JavaScript space.

My curiosity stems mostly from the fact that things like CodeMirror and ACE have been available in the JavaScript universe for a while (and LightTable, Atom, etc. leverage those kinds of "surfaces") and yet nobody seems to have tried to get, say, node-webkit to run something like acme.

So 1) all that's really necessary is the smart surface, 2) there are known instances of said surfaces (and plenty of people using them for a new generation of editors) and yet 3) nobody seems to have thought of reimplementing acme atop them.