How to Compile and Run
To compile RealVoice, on an SGI Indigo2, type the command:

% make

RealVoice can be run in two modes, one which accepts connections, and one which attempts to make one. To run RealVoice to accept connections, type the command:

% RV

To run RealVoice to attempt to make a connection, type the command:

% RV USER_NAME MACHINE_NAME

Where USER_NAME is the name of the user you wish to connect to, and MACHINE_NAME is the machine you wish to connect to.

NOTE: REALVOICE MAKES USE OF SILICON GRAPHICS SOUND INPUT AND OUTPUT LIBRARIES. YOU SHOULD BE ON A WORKSTATION WITH SPEAKERS (OR HEADPHONES) AND A MICROPHONE.

ALSO NOTE: A BINARY VERSION OF REALVOICE HAS BEEN INCLUDED IN CASE THE ABOVE MAKE COMMAND DOESN'T WORK ON A SPECIFIC SYSTEM. THE EXECUTABLE IS GZIP COMPRESSED. THE FILE IS RV.GZ. ONCE UNZIPPED, THIS SHOULD RUN ON ANY MODERN SGI SYSTEM.

Overview
RealVoice is a network protocol that provides for real-time transmission of voice data across the internet between two peers. Essentially, then, RealVoice creates telephone-call semantics for voice conversations over the net.As a high-level protocol, RealVoice includes different levels of protocol definition.

• Network-Level Protocol: This protocol level defines the way RealVoice connections are made and maintained, how data is sent across the net (packet header formats, etc.), and what happens in exceptional circumstances. This protocol is at the TCP and UDP level (Writeup by Timothy Milliron)
• Data-Level Protocol: Assuming that well-defined methods of connecting and transporting data across the net exist, we define protocols for how latency and low bandwidth are tolerated in transporting voice data (a low latency and high bandwidth commodity), including compression and interpolation techniques. (Writeup by Philip Nikolov)
• User-Level API: If the protocol were to be used in any real way outside an academic setting, a convenient user-level programming interface would be needed to easily access the system's functionality. In fact, such a user-level protocol is defined by the RealVoice and RealVoiceCallback classes, and by the RealVoiceUtils.H file.

RealVoice also refers to the application RV, which participates in the protocol defined by RealVoice, and also provides the basic functionality of playing and receiving sound data. The RealVoice sound input and sound output classes are designed to provide a system-independent interface to system-dependent routines and system calls, and are implemented on SGI systems (writeup by Boris Kerbikov)

Platforms
The RealVoice protocol layers, and the RV application that receives and plays the sound data, are all implemented on the Silicon Graphics IRIX platform. Network and Sound input and output classes are designed to be easy to port to other systems, though no such port was made.

Work Distribution

• Boris Kerbikov wrote the SoundInput and SoundOutput classes, and did a lot of work developing threads to read and write sound data in the background (though we didn't end up using any of them)
• Timothy Milliron wrote a full-featured sockets library (TCPSocket, TCPServerSocket, UDPSocket classes), including non-standard classes (UDPOrderedSocket) for specific use with RealVoice, and also implemented most of the protocol-handling code (RVConnectionHandler, RealVoice)
• Philip Nikolov investigated and implemented strategies and protocols to deal with latency tolerance and bandwidth limitation for voice transmission.