R  = request
A  = acknowledgement (response)
[] = optional TLV

   client                                server
   ------                                ------

Starting execution of a process:

1. machine_process_execute(R)
     - TLV_TYPE_MACHINE_PROCESS_PATH
	  - [TLV_TYPE_MACHINE_PROCESS_FLAGS]
	    - Hidden, etc
                                         machine_process_execute(A)
														 - TLV_TYPE_RESULT
													    - [TLV_TYPE_CHANNEL_ID]
 
Receiving output from the process:

2.                                       core_channel_write(R)
                                           - TLV_TYPE_CHANNEL_DATA_GROUP
														   - TLV_TYPE_CHANNEL_ID
															- TLV_TYPE_CHANNEL_OUTPUT_DATA


How channels should work on a theoretical level:

1) Channels are logical communication units with two endpoints.  As such, a 
   channel has four logical operations:

	a) Open

		Opening a channel involves two sides creating a logical 'connection' 
		and agreeing on a common identifier for the channel.  This channel will
		henceforth be referred to as the ``channel identifier''.

	b) Read

		Reading from a channel on either endpoint simply means that data is being
		read from the 'output' buffer on the respective endpoint.  When a client
		reads from a channel that is associated with a process, they are simply
		reading data that has been written to standard output.  When a client reads
		from a channel that is associated with a file, they are simply reading 
		data from the file itself at the current offset.

	c) Write

		Writing to a channel from either endpoint simply means that data is being
		pushed into the 'output' buffer for the respective endpoint.  When a client
		writes to a channel that is associated with a process, they are, in effect,
		writing to the process.  When a client writes to a channel that is associated
		with a file, they are simply writing to the file at its current offset.

	d) Close

		The close operation simply tears down the previously established 'connection'.


	Channel endpoints will be required to implement handlers for each of the 
	aforementioned operations.  The read and write operations should have 
	substitutable handlers for allowing for extended reading and writing methods 
	depending on the underlying endpoint.