I love your idea of having a widget that represents "the window", but
I'm not too keen on the gestures; the rest of our desktop has no
reliance on the things, aiming more for physical, direct interaction.
Yes, I know it's a change of concept, but not necessarily a big one. In fact, it's in some way more "physical" dragging something up to maximize the window and down to minimize than pushing a button to make the same action. In the end, if we change our assumed ideas, it's really more natural and intuitive. We'd just need to learn a new way of doing things.
We could enhance that design by having it work
within a workspace, too, so you don't need to use the overlay mode to
add a workspace visually; dragging the Esfera widget would zoom out,
revealing hotspots on the screen to perform different interactions.
(Could even be the same list you describe).
So, if I've understood well, what you propose is that, when dragging the widget, some hotspots will appear around it, and dragging it to them would produce different actions, right?
This is an interesting idea. It'd be easier to use by new users, who will see more clearly the possible actions to do, and we'd avoid problems caused by poor gesture recognition. But, as a disadvantage, the actions would be slower to perform than if we use gestures; we would loose the intuitiveness and speed that the gestures offer. Anyway, I think it's something to think about.
But I deffinitely wouldn't wait until the user releases the widget to do the action; this would make it slower and feel less useful. I'd perform the action before he releases it, and if it's unwanted or he changes his mind, he just needs to take the widget back to undo it. This way it would be very quick and responsive. Using gestures, the action would be performed as soon as a gesture is recognized; using hotspots, it would be performed as soon as the hotspot is reached, but they should be close to the widget to make it fast.
At that point, it may be interesting to improve drag / drop in
general. I believe the system can be aware of any droppable areas for
the current drag operation, so why do we torture our users by forcing
them to drag a widget all that way there over a 38" 4k pixel monitor?
Snapping or physics (think kinetic scroll, with magnetic hot zones)
could be worth some experimentation. (All things considered, the end
result of that would basically feel like gestures, but be more
wholesome interactions).
This seems an amazing idea to explore!