Networks
A computer network links devices so they can exchange data, over copper cable, radio (wireless LANs) or optical fibre. Data travels as packets, each carrying addresses, and is handled by a stack of protocol layers from the physical link up to applications. Routers join separate networks together by forwarding each packet on its own, while the end devices keep track of reliability.
Packet switching and statelessness
Packet switching (theory first published by Kleinrock in 1961, developed in parallel at RAND and the UK's NPL) shares link capacity among many flows. In the Internet architecture (RFC 1122), routers are stateless: each IP datagram is forwarded independently, so redundant paths can be used when a router or link fails.
- Link-layer frames carry IP datagrams one hop at a time; link addresses (EUI-48) are only meaningful on the local link.
- IP addresses are end-to-end identifiers used for routing between networks.
- End-to-end reliability, ordering and flow control are implemented by hosts (e.g. TCP), not routers.
Attenuation in decibels; 95% transmission over 1 km is about 0.22 dB/km (our calculation).
Full explanation — the complete reference version every reading depth is based on
What a network is
A network is a set of devices — computers, phones, printers, sensors — connected so they can send data to one another. Links may be wired, as in Ethernet (standardised by the IEEE 802.3 Working Group), wireless, as in wireless local area networks (IEEE 802.11, sold under the Wi-Fi name), or optical fibre, which carries data as light.
Packets instead of a reserved line
Rather than reserving a continuous connection for each conversation, data networks split messages into packets that share the links with everyone else's packets. Leonard Kleinrock published the first paper on packet-switching theory in July 1961. Similar work went on at the same time, independently, at RAND in the USA and at the National Physical Laboratory in the UK, where the word 'packet' came from. These ideas became the basis of the ARPANET and then the Internet.
Layers
- Application layer: the programs' own protocols, such as web, email or name lookup.
- Transport layer: end-to-end delivery between programs, for example TCP (reliable) or UDP (connectionless).
- Internet layer: the Internet Protocol (IP), which moves datagrams from source host to destination host across networks.
- Link layer: the protocol for the directly connected network, such as Ethernet or a wireless LAN.
Each layer relies on the one below and offers a service to the one above, so a web browser does not need to know whether its data travels over radio, copper or glass.
Addresses and routers
- Network hardware interfaces usually carry a 48-bit identifier (EUI-48, often called a MAC address) used on the local link.
- IP addresses identify hosts across interconnected networks.
- Routers (called 'gateways' in the original specifications) join networks and forward each IP datagram independently of the others.
- Reliability and flow control are handled end to end by the hosts, not by the routers in between.
Worked example: light in a fibre
Before Kao's work in the 1960s, only 1% of the light entering a glass fibre survived 20 metres. Charles Kao and George Hockham aimed for at least 1% to remain after 1 km (Kao presented their conclusions in 1966). By 2009 about 95% remained after 1 km. If each kilometre keeps 95%, then after 3 km 0.95 × 0.95 × 0.95 ≈ 0.857, so about 86% remains (our calculation) — which is why long fibre links still need optical amplifiers along the way.
Power remaining after L kilometres if each kilometre keeps 95% of the light (illustrative model).
How we know
The layered architecture and the stateless-router principle are written down in the Internet's own standards (RFC 1122, 1989). The history comes from 'A Brief History of the Internet', written by people who built it, including Kleinrock himself. The fibre figures come from the Nobel Prize committee's account of Kao's work.
Assumptions and limits
The four-layer picture is a design discipline: real devices sometimes cross layers, and middleboxes such as firewalls and address translators do keep per-connection state. The 95%-per-kilometre model ignores splices, connectors and dispersion, which also limit real links.
Ask ScienceVerse
Still curious about Networks? Ask a question, get hints, take a short lesson or try a challenge. The tutor answers only from this concept's approved sources, and says so when it has none.
Ask the tutor about this concept on the full tutor page.
Connections
Guided learning path
See everything to learn before this, in order, with your progress:
Check your understanding
Take a quick check of two to five questions, with an explanation for every answer:
Sources and methodology
- Leonard Kleinrock published the first paper on packet-switching theory in July 1961 and the first book on the subject in 1964. (awaiting scientific review)
- Early packet-switching work at MIT (1961–1967), RAND (1962–1965) and the UK's National Physical Laboratory (1964–1967) proceeded in parallel, and the word 'packet' was adopted from the NPL work. (awaiting scientific review)
- The Internet architecture organises a host's communication protocols into layers — application, transport, internet and link — and a host typically must implement at least one protocol from each layer. (awaiting scientific review)
- RFC 1122: Requirements for Internet Hosts — Communication Layers — Government or standards body
- To communicate on its directly connected network, a host must implement the link-layer (media-access) protocol used by that type of network, and there is a wide variety of such protocols for different network types. (awaiting scientific review)
- RFC 1122: Requirements for Internet Hosts — Communication Layers — Government or standards body
- In the Internet architecture, networks are interconnected by packet-switching computers called gateways or IP routers, which forward each IP datagram independently of other datagrams. (awaiting scientific review)
- RFC 1122: Requirements for Internet Hosts — Communication Layers — Government or standards body
- In the Internet architecture, the state needed for end-to-end reliability and flow control is kept in the communicating hosts (in the transport layer or applications), not in the routers between them. (awaiting scientific review)
- RFC 1122: Requirements for Internet Hosts — Communication Layers — Government or standards body
- The IEEE 802.3 Working Group develops the standards for Ethernet networks. (awaiting scientific review)
- IEEE 802.3 Ethernet Working Group — Government or standards body
- The IEEE 802.11 Working Group develops the standards for wireless local area networks (WLANs). (awaiting scientific review)
- IEEE 802.11 Working Group for Wireless Local Area Networks — Government or standards body
- A 48-bit Extended Unique Identifier (EUI-48) is commonly used as the address of a hardware interface under IEEE 802 standards, where such globally unique addresses are sometimes called MAC addresses. (awaiting scientific review)
- Guidelines for Use of Extended Unique Identifier (EUI), Organizationally Unique Identifier (OUI), and Company ID (CID) — Government or standards body
- Before Kao's work in the 1960s, only 1 percent of the light entering a glass fibre remained after just 20 metres; Charles Kao and George Hockham set the goal that at least 1 percent should remain after 1 kilometre, and Kao presented their conclusions in January 1966. (awaiting scientific review)
- By 2009, high-purity optical fibre let about 95 percent of the light remain after travelling one kilometre, and over long distances the signal is reinforced by optical amplifiers. (awaiting scientific review)
Claims marked “awaiting scientific review” cite the sources listed but have not yet been signed off by a scientific reviewer.
Content status: published 1 October 2026.
- Scientific review: this version has not yet been signed off by a scientific reviewer.
- The Advanced explanation has not yet been reviewed for age suitability.