My name is Evan Bolyen and I would really love to work with you all on
I am still reading through your coffeescript so I apologize if I have
misunderstood anything, please feel free to correct me.
I was talking to Bryan at SciPy and he mentioned how different frames(?)
will occasionally fail because they are drawn before Kiwi is finished.
I have explored this particular situation before and a good solution I
have found is to explicitly coordinate reactive variables. This has a
can be implemented as an observer pattern over a single variable. Then,
different components can choose to listen to this while providing a
continuation. This provides loose coupling and complete control over
when an action will be preformed without resorting to some kind of
global state or synchronization scheme over an event bus.
Suppose each frame provided a reactive property called "ready". A single
coordinator will register itself to each frame's "ready"
property/channel, where the continuation it provides will modify the
coordinator's internal state. When all (or enough) frames are ready, the
coordinator can produce an action over the frames to begin rendering.
This can be extended throughout the architecture to potentially produce
a "cascade" of state transitions that can be controlled and reasoned about.
What are your thoughts on this strategy?