Packet switching

Computers often generate bursts of data rather than continuous traffic. Packet switching divides communication into smaller units carrying control information for identification and delivery. Network devices can interleave packets from different exchanges, sharing links according to arriving work instead of reserving the whole transmission path exclusively for each conversation.

References: [1]

An intermediate device selects an outgoing link for each packet according to forwarding rules. Busy links can cause queuing, and insufficient resources can lead to loss. Delivery time reflects packet size, link rate, processing, and waiting, so sharing transmission capacity also creates the need to manage contention and congestion.

References: [2]

Packets can experience different delays or paths and may arrive out of order. Reassembly, loss detection, and retransmission depend on the protocols in use. IP provides best-effort datagrams, while transports such as TCP add ordered byte-stream behavior. The general switching approach does not prescribe one universal recovery procedure for every application.

References: [1]

Early packet networks combined theoretical work with interface equipment and live link experiments. ARPANET connected heterogeneous hosts through interface message processors. Later internetworking addressed differences among complete networks, using common datagrams and interfaces so packets could cross those boundaries while local technologies retained room to evolve independently.

References: [2]