In store-and-forward switching, a question arises concerning the size of what we are switching. The earliest electronic store-and-forward network was the telegraph. In the telegraph, the entirety of a user’s message (called a telegram) was switched through the telegraph office at once. Most of these messages were very small because the telegraph office charged by the letter or word. As messages get larger, message switching is unwieldy. Given the size of computer files today, message switching is unacceptable. Modern networks carry packets. A packet is a relatively small transmission unit with an upper bound. Each packet contains its own addressing information. To send a large book, for example, the transmitting system would break the book into small chunks, each of which is probably the equivalent of a page of text. The individual pages (or packets) are sent across the communication network and reassembled at the far end. Very short messages, however, may be entirely contained in a single packet. With this kind of structure, other traffic can be inserted between the pages of the book. And if anything causes a packet to be destroyed or lost, only that single packet (or perhaps a small number of packets) must be retransmitted, not the entire book.
This brings us, then, to the definition of packet switching. A packet switched network is one that employs store-and-forward switches interconnected by statistically time division multiplexed (STDM) facilities, and uses this network to pass individually addressed packets from one end-system to another.