rustyforks/screenshare

Screensharing via WebRTC

Screenshare

Screensharing via WebRTC from a single producer to web client consumer's. The initial communication between the producer and a consumer is done via the signaling component over websockets. Once the producer and consumer have decided how to send the video stream between each other, they will establish a direct connection and send the video stream over that connection.

Message format

The WebRTC signaling message are sent as json. Format of messages:

{
    "action" : "<ACTION>",
    "data": "<DATA>"
}

The data field can hold arbitrary data and its format differs depending on the type of message that is being set. The action field represents the type of the message that is being sent. List of valid action values:

• answer
• offer
• candidate
• response

The three first items in the list represents messages that are sent from hosts to signaling server. The last value is used by the signaling server when responding to messages from a host.

Messages sent from the signaling server to the hosts might have some extra optional fields depending on the message that is being sent:

• The producer communicates over a single websocket connection to the signaling server, so the signaling server needs to add extra information in the messages so that the producer knows which consumer sent the message. This is handled in the signaling server by adding a extra id field in the messages sent to the producer. This id field contains a unique identifier for a specific client.
• The signaling server sets the action field to response when responding to the hosts, a extra type field is added that stores the original action value set by the host.

Components

The application is divided into three modular components. The components communicate with each other over websockets and WebRTC.

Signaling

The signaling component handles the WebRTC signaling between the producer and consumer's. The signaling logic is designed to be hosted in AWS. It consist of multiple Java modules where every module represents a Lambda function. The name of the child module represents the route key that should be used in the API gateway that calls the Lambda function.

When setting up a websocket connection via AWS API Gateway, a consumer and producer are differentiated by the type parameter in the URL. So when connecting as a producer to the signaling server, the type parameter should be set to producer in the URL like so:

wss://<API_ID>.execute-api.<REGION>.amazonaws.com/<STAGE>?type=producer

A consumer doesn't need to specify any parameter since a empty type parameter is assumed to represent a consumer.

Requirements

• Java (Version known to work: 8)
• Maven (Version known to work: 3.6.1)

Build

The following command should be ran in the parent/aggregate folder:

mvn clean install

Install

The jars generated during build should be uploaded as the code for AWS Lambda functions.

Producer

The producer produces the video stream and shares its screen. The implementation is written in rust and uses ffmpeg to record the screen.

Build

cargo build --release

Run

USAGE:
    screenshare.exe [OPTIONS] --url <URL>

OPTIONS:
    -c, --config <cfg>    Path to the config file containing the arguments to the FFMPEG command.
                          [default: ffmpeg.cfg]
    -h, --help            Print help information
    -s, --stun <stun>     The URL of the stun server which we should use when setting up the RTC
                          connections. If not set, a default google stun server will be used.
                          [default: stun:stun.l.google.com:19302]
    -u, --url <URL>       The URL of the websocket server which we will connect to.

Requirements

• Rust (Version know to work: 1.57.0)
• Cargo (Version know to work: 1.57.0)
• ffmpeg (Version known to work: 4.4.1)

For the default settings stored in ffmpeg.cfg, ffmpeg must have support for x264 and gdigrab (windows only). But these settings can be changed at runtime, so the only real requirement is a working ffmpeg binary.

Consumer

A webclient written in Javascript. The consumer is expected to send the WebRTC offer while the producer sends answer's.

Run

Specify the domain name of the signaling server and click connect. The video stream from the producer should soon start playing in the HTML video element of the client.