And:  The "hardware" outputs from Sonar are actually Sonar's final output(s). Although they are connected to hardware/speakers for most of us, they are not directly connected to hardware.
 
For me Sonar's outputs feed track strips in the Patchmix program, which can be routed to other asio pipes/conduits/streams, and/or to actual hardware outputs.
 
The CW dudes should change this nomenclature to reflect current reality. Don't ya think?
 
 

Yo John
 
I understand the historical consensus.
 
My first bus is for sound level adjustment.  It is NOT for a project's final mix.  I never touch or see Sonar's final output buses, in the third panel.
 
The 2nd bus is the final mix bus.
 
As the diagram shows: when a mix, from the master mix bus, is bounced to tracks, the output of the new track is sent to the Main Out BUS.  This maintains levels for the speakers, and other monitoring tools.
 
This is in keeping with using a control surface as you have described for a  "Master" bus.
 
My difference is the master bus and the master projects mixing bus are NOT the same.
 
 
 

I suppose it depends on what you mean by connected. In no case are they directly connected to hardware. There has to be a driver involved. 
 
The point is that they are the last thing as far as Sonar is concerned for outputing audio. 

This is the flow of the signal:
Tracks -> Buses -> Main outs

 
+1
 
That's the short description.
 
 
Tracks  (send)->   n..(Bus/out)
           (post fader)-> Bus/Out
 
Bus    (send) ->    n..(Bus/Out)
          (post fader)  ->     Bus/Out
 
Out -> driver->Hardware
            
The big deal is in terminology, and software mixers give you way more rope to hang yourself with than a hardware mixer ever did. Hardware mixers had fixed inputs, inserts, buses, and mains - the architecture and routing capabilities are set in stone. Here in Sonar, there are no fixed limitations. If you want to put reverb on every channel you can certainly do so if you have the horsepower - and skip the bus altogether. A bus is not required, although it's a convenience that makes sense.
 
Personally, my endpoint is a bus I call Master all other buses may feed another bus, either through a send or post fader to another bus, but eventually the signal all flows to Master. Master feeds a Mains bus and out to the speaker. 
 
When exporting you have a choice of methods. You can choose 'entire mix' which would be from your mains. or you could export buses and select 'Master' (in my case).
 
Sonar gives you the rope - it's up to you how you use it. There is no routing that is perfect for all, although there are rules of thumb that work as guidelines (but are not mandates).