WebSockets API: Writing WebSocket server

Introduction

If you would like to use the WebSocket API, it is useful if you have a server. In this article I will show you how to write one in C#. You can do it in any server-side language, but to keep things simple and more understandable, I chose Microsoft's language.

This server conforms to RFC 6455 so it will only handle connections from Chrome version 16, Firefox 11, IE 10 and over.

First steps

WebSockets communicate over a TCP (Transmission Control Protocol) connection. Luckily, C# has a TcpListener class which does as the name suggests. It is in the System.Net.Sockets namespace.

TcpListener

Constructor:

TcpListener(System.Net.IPAddress localaddr, int port)

localaddr specifies the IP of the listener, and port specifies the port.

Methods:

• Start()
• System.Net.Sockets.TcpClient AcceptTcpClient()
  Waits for a Tcp connection, accepts it and returns it as a TcpClient object.

Here's a barebones server implementation:

​using System.Net.Sockets;
using System.Net;
using System;

class Server {
    public static void Main() {
        TcpListener server = new TcpListener(IPAddress.Parse("127.0.0.1"), 80);

        server.Start();
        Console.WriteLine("Server has started on 127.0.0.1:80.{0}Waiting for a connection...", Environment.NewLine);

        TcpClient client = server.AcceptTcpClient();

        Console.WriteLine("A client connected.");
    }
}

TcpClient

Methods:

• System.Net.Sockets.NetworkStream GetStream()
  Gets the stream which is the communication channel. Both sides of the channel have reading and writing capability.

Properties:

• int Available
  This Property indicates how many bytes of data have been sent. The value is zero until NetworkStream.DataAvailable is true.

NetworkStream

Methods:

• Write(Byte[] buffer, int offset, int size)
  Writes bytes from buffer, offset and size determine length of message.
• Read(Byte[] buffer, int offset, int size)
  Reads bytes to buffer. offset and size determine the length of the message.

Let us extend our example.

TcpClient client = server.AcceptTcpClient();

Console.WriteLine("A client connected.");

NetworkStream stream = client.GetStream();

//enter to an infinite cycle to be able to handle every change in stream
while (true) {
    while (!stream.DataAvailable);

    Byte[] bytes = new Byte[client.Available];

    stream.Read(bytes, 0, bytes.Length);
}

Handshaking

When a client connects to a server, it sends a GET request to upgrade the connection to a WebSocket from a simple HTTP request. This is known as handshaking.

This sample code can detect a GET from the client. Note that this will block until the first 3 bytes of a message are available. Alternative solutions should be investigated for production environments.

using System.Text;
using System.Text.RegularExpressions;

while(client.Available < 3)
{
   // wait for enough bytes to be available
}

Byte[] bytes = new Byte[client.Available];

stream.Read(bytes, 0, bytes.Length);

//translate bytes of request to string
String data = Encoding.UTF8.GetString(bytes);

if (Regex.IsMatch(data, "^GET")) {

} else {

}

The response is easy to build, but might be a little bit difficult to understand. The full explanation of the Server handshake can be found in RFC 6455, section 4.2.2. For our purposes, we'll just build a simple response.

You must:

1. Obtain the value of the "Sec-WebSocket-Key" request header without any leading or trailing whitespace
2. Concatenate it with "258EAFA5-E914-47DA-95CA-C5AB0DC85B11" (a special GUID specified by RFC 6455)
3. Compute SHA-1 and Base64 hash of the new value
4. Write the hash back as the value of "Sec-WebSocket-Accept" response header in an HTTP response

if (new System.Text.RegularExpressions.Regex("^GET").IsMatch(data))
{
    const string eol = "\r\n"; // HTTP/1.1 defines the sequence CR LF as the end-of-line marker

    Byte[] response = Encoding.UTF8.GetBytes("HTTP/1.1 101 Switching Protocols" + eol
        + "Connection: Upgrade" + eol
        + "Upgrade: websocket" + eol
        + "Sec-WebSocket-Accept: " + Convert.ToBase64String(
            System.Security.Cryptography.SHA1.Create().ComputeHash(
                Encoding.UTF8.GetBytes(
                    new System.Text.RegularExpressions.Regex("Sec-WebSocket-Key: (.*)").Match(data).Groups[1].Value.Trim() + "258EAFA5-E914-47DA-95CA-C5AB0DC85B11"
                )
            )
        ) + eol
        + eol);

    stream.Write(response, 0, response.Length);
}

Decoding messages

After a successful handshake, the client will send encoded messages to the server.

If we send "MDN", we get these bytes:

Let's take a look at what these bytes mean.

The first byte, which currently has a value of 129, is a bitfield that breaks down as such:

• FIN bit: This bit indicates whether the full message has been sent from the client. Messages may be sent in frames, but for now we will keep things simple.
• RSV1, RSV2, RSV3: These bits must be 0 unless an extension is negotiated which supplies a nonzero value to them.
• Opcode: These bits describe the type of message received. Opcode 0x1 means this is a text message. Full list of Opcodes

The second byte, which currently has a value of 131, is another bitfield that breaks down as such:

• MASK bit: Defines whether the "Payload data" is masked. If set to 1, a masking key is present in Masking-Key, and this is used to unmask the "Payload data". All messages from the client to the server have this bit set.
• Payload Length: If this value is between 0 and 125, then it is the length of message. If it is 126, the following 2 bytes (16-bit unsigned integer) are the length. If it is 127, the following 8 bytes (64-bit unsigned integer) are the length.

Note that the MASK bit is set in our message. This means that the next four bytes (61, 84, 35, and 6) are the mask bytes used to decode the message. These bytes change with every message.

The remaining bytes are the encoded message payload.

Decoding algorithm

D_i = E_i XOR M_{(i mod 4)}

where D is the decoded message array, E is the encoded message array, M is the mask byte array, and i is the index of the message byte to decode.

Example in C#:

Byte[] decoded = new Byte[3];
Byte[] encoded = new Byte[3] {112, 16, 109};
Byte[] mask = new Byte[4] {61, 84, 35, 6};

for (int i = 0; i < encoded.Length; i++) {
    decoded[i] = (Byte)(encoded[i] ^ mask[i % 4]);
}

Related

• Writing WebSocket servers