Gadgets could use a re-examination

Summary: See Details, but basically it's time to step back and think about them.

Time for another round of stepping back, looking at Gadgets, and thinking about how to make them work better. Where are things too ad-hoc, too boilerplate, too fragile? Can we make the behavior more declarative, more based on simply declaring relationships between data and rendering? Where are common patterns to lift out? Do we need the vdom idea, or can we stay away from it? I believe that the paradigm I’m drifting into is reactive-databound, where a Gadget’s rendering should consist of some Rx input state, combined with some Rx local state?