Notes for Team 2

(From September 25 meeting with Prof. Keshav)

There are two control levels - the application level and the basic one. You are to implement the basic level: setting up the call, tearing down the call, etc.
Use SIP. It has many parts, including directory service. You should concentrate only on the signalling part. Look at Henning's paper; it has a good description of SIP.
Basic SIP is just between two endpoints. But you will have a gateway, so you must implement signalling. Start with the invite-ack step. When a call is initiated, the PBX rings and picks up. The signal comes in on one of the 4-8 lines to the gateway. Look up the proper IP address (from directory services) for this phone call, and use the gateway to send an invite message to the PC.
The PC responds to the gateway, who can then send a JTAPI message to the PBX. The only parts of JTAPI you need are core and CallControl.
Allow for extension numbers. The signalling team implements the "please enter extension" message, and then collects the DMTF tones as the caller presses the keypad.
Handling unforeseen events is the hard part. One of your goals is to figure out what it is so hard to deal with the uncommon cases.

Is this team responsible for the data? No. Coordinate with Team 1. Data will be in RTP format.
What happens when the phone hangs up? The PBX gets the signal, sends a protocol message to the gateway saying this has happened, who then breaks the association with the PC, and makes the line a now-available resource.
What if the PC goes down? This is harder. The data team has to notice no traffic after a while and tell you. You can then send another invite. If no ack in a reasonable time, tear down the call.
Must we allow conferencing? SIP allows you to join multi-party calls, so yes, you should allow conferencing.
Do the signalling software and the client run on the same machine? Typically not. The directory service gives you the current IP number and port of the client. The directory is a "location server".