Oh 
my bad then @hauleth, I misread that, sorry!
Another obvious point not discussed is the potential cognitive savings and just overall simplification of using a single transport. I think most apps will grow beyond request/response cycle and require some realtime functionality at some point. When that point is reached, you’re now managing both HTTP and websockets.
Since websockets obviously also support request/response, you now have 2 places where you can place your request handlers. Why would you pick a controller over a channel if you already have a channel set up?
