Synthesizer issues

Is VC from A to B same as from B to A?  No

Is gateway/transcoder on the end hosts or always a point between?
assuming it's going to be a capability on one of the end hosts or smart routers

H.323 standard claims to support IP multicast. 

write code to identify need for transcoder

not assuming any discovery mechanisms...all information is provided in user request

are these peer to peer conferencing or are there special servers?  hmmm...

Are there synthesizers at every edge router?  Admin network?
assuming just one synthesizer that coordinates everything.  

scope
Assuming intranet set up for now...for full deployment - it will be on parts of internet.

*obtaining an IP multicast address is responsibility of the component
 
Components are all throughout the internet and control/administer a particular 'subnet'?

JC says H.323 support of multicast too open for interpretation.  

synthesizer - asynchronous?  multi-threaded needed?
assuming single thread for now

throwing out HTML web form idea...there will be services that will take tens of seconds for a response...

**should all services be available via a IP/port pair?
do/should components share a similar interface?

SIP component will stop after one error...future goal may be for it to 
communicate multiple errors in response to one request.  

is Synthesizer meant to store state?  Or does it just initiate and exit?
assuming latter.

all components comminicate to synthesizer via a socket interface

according to YangHua' paper - finding an overlay for EndSysMult takes 4 MINUTES without prior network info!

need to define differences between setup and execution.  


Things learned during prototype:

Components will have different response times.  
	Jiacheng's gateway -> 30 sec
	YangHua's/Sanjay's Narada overlay path -> 4 minutes

Current inter-component communication:
	sockets
	future: XML?

SIP gateway reporting one error and exiting.  

Synthesizer definitions:
	Should only handle setup - should not need to store state

Questions for component interactions:
	How are these recipes defined and loaded?
	Which component to call first?
	Can user input change order? (i.e. Narada before IP multi)

Assumptions made so far:
	All information is provided from user request and all services available -  no discovery mechanisms
	
temporary solution - in the input file specify whether to use end sys multicast
if don't care, use IP multicast - if error then use end sys multicast.

Is it really possible to have a generic component1->synth->component2 structure?

establish - user client should not need to know if IP multicast or ESM is being used.  

can ESM get assigned the IP multicast address to listen to?


assumptions:
IP addresses from synth to vid gateway is for SIP only...yup

what needs to be same?  multicast IP for proxy and vic, 
port numbers for control and data
	
problem...we need to start proxy...proxies all need to be speaking on same port