diff --git a/Client-server-discovery-lifecycle.md b/Client-server-discovery-lifecycle.md
index 2decb17..ec019f5 100644
--- a/Client-server-discovery-lifecycle.md
+++ b/Client-server-discovery-lifecycle.md
@@ -105,8 +105,10 @@ The interface has two methods.  One responsible for starting the Server, and ano
 
 1. Obtaining the lock file for the CWD + AppConfiguratoin-hash pairing.  (That is there can only be one server of a given configuration (organization + module id) in any working directory).
 2. Reading the ServerPorts from the active configuration.
-3. Checking to see if the old server is alive.  If so, report these ports to the client (TODO - how?)
-4. If the old server was not alive, start up a new server (fork + detach?).  Write the new server bindings to the lock file.
+3. Instantiating an instance of the ServerMain class denoted in the configuration.
+3. Checking to see if the old server is alive using the ServerMain instance.  If so, report these ports to the client (TODO - how?)
+4. If the old server was not alive, start up a new server using the ServerMain instance.  Write the new server bindings to the lock file and release the lock.
+5. Notify the client (via stdout) what the bound ports are.
 
 Port bindings come in java.net.URI, which includes an address, a protocol and a port.  The server will actually also write down the PID (process ID) of the server (assuming it can obtain the pid).  This enables clients to "kill rogue sbt server" directly.