How do you think this compares to Perl::Mason? How are you avoiding the same pitfalls that Mason has? complicated failure modes, pretty much impossible to improve timing and create on demand resource prefetching without increasing complexity greatly?
I don't know much about Mason, but this isn't a templating engine; it's just a wrapper library on top of the DOM API. Since it doesn't need any compilation or coordination outside of the JavaScript process, it's much less complex, and composes well without passing partials and other cruft into the template data.