Transmission Control Protocol

When two computers wish to exchange information over a network, there are several components that must be in place before the data can actually be sent and received. Of course, the physical hardware must exist, which is typically either a network interface card (NIC) or a serial communications port for dial-up networking connections. Beyond this physical connection, however, computers also need to use a protocol which defines the parameters of the communication between them. In short, a protocol defines the "rules of the road" that each computer must follow so that all of the systems in the network can exchange data. One of the most popular protocols in use today is TCP/IP, which stands for Transmission Control Protocol/Internet Protocol.

By convention, TCP/IP is used to refer to a suite of protocols, all based on the Internet Protocol (IP). Unlike a single local network, where every system is directly connected to each other, an internet is a collection of networks, combined into a single, virtual network. The Internet Protocol provides the means by which any system on any network can communicate with another as easily as if they were on the same physical network. Each system, commonly referred to as a host, is assigned a numeric value which can be used to identify it over the network. These numeric values are known as IP addresses, and are usually represented as a string value that contains a series of numbers.

There are two versions of TCP/IP and two different IP address formats based on which version of the protocol is being used. For Internet Protocol v4 (IPv4), addresses are 32 bits wide and are represented by a sequence of four 8-bit numbers separated by periods. This is called dot-notation and looks something like 192.168.19.64. This is the address format that many developers are familiar with because IPv4 continues to be the most commonly used version of the protocol. Internet Protocol v6 (IPv6) is the next generation of IP and it supports a much larger address space as well as a number of other features. IPv6 addresses are 128 bits wide and represented by a sequence of hexadecimal values separated by colons. As expected, this format is much longer than the simple dot-notation used by IPv4 address. A typical IPv6 address will look something like fd7c:2f6a:4f4f:ba34::a32, although there are certain shorthand notations that can be used. SocketTools supports both IPv4 and IPv6, and can automatically determine which version of the protocol should be used based on the address. Because IPv4 is still widely used, if given a choice between using IPv4 or IPv6, the SocketTools components will choose IPv4 for backwards compatibility whenever possible. However, an application can choose to exclusively use IPv6 if required.

When a system sends data over the network using the Internet Protocol, it is sent in discrete units called datagrams, also commonly referred to as packets. A datagram consists of a header followed by application-defined data. The header contains the addressing information which is used to deliver the datagram to its destination, much like an envelope is used to address and contain postal mail. And like postal mail, there is no guarantee that a datagram will actually arrive at its destination. In fact, datagrams may be lost, duplicated or delivered out of order during their travels over the network. Needless to say, this kind of unreliability can cause a lot of problems for software developers. What's really needed is a reliable, straightforward way to exchange data without having to worry about lost packets or mixed data.

To fill this need, the Transmission Control Protocol (TCP) was developed. Built on top of IP, TCP offers a reliable, full-duplex byte stream which may be read and written to in a fashion similar to reading and writing a file. The advantages to this are obvious: the application programmer doesn't need to write code to handle dropped or out-of-order datagrams, and instead can focus on the application itself. And because the data is presented as a stream of bytes, existing code can be easily adopted and modified to use TCP.

TCP is known as a connection-oriented protocol. In other words, before two programs can begin to exchange data they must establish a connection with each other. This is done with a three-way handshake in which both sides exchange packets and establishes the initial packet sequence numbers. The sequence number is important because, as mentioned above, datagrams can arrive out of order; this number is used to ensure that data is received in the order that it was sent. When establishing a connection, one program must assume the role of the client, and the other the server. The client is responsible for initiating the connection, while the server's responsibility is to wait, listen and respond to incoming connections. Once the connection has been established, both sides may send and receive data until the connection is terminated.

Most of the application protocols which are supported by SocketTools use TCP to communicate over the Internet or local intranet. However, it is important to remember that it is not necessary for you to understand how TCP/IP works at the lowest levels in order to use SocketTools. Complex operations such as performing checksums on packets of data to ensure they arrive intact are handled for you automatically. In most cases, the SocketTools interface provides methods which are similar to what you would use when reading or writing to a file.