4.9.3.1 - The Internet, packets, routing and DNS

Networks, routers and gateways

The Internet is a worldwide interconnection of networks that communicate using the Internet protocol suite.

Individual local networks connect through Internet service providers and larger interconnected networks. Links and routers form many possible paths. There is no single central computer through which every Internet exchange must pass. Different organisations operate different parts while common protocols let them exchange traffic.

A router has connections to multiple networks and chooses an outgoing connection or next hop. A school router can forward traffic for an external address to its provider while keeping traffic for a directly connected network on that network. A router does not normally need to know the complete path to the destination; it needs a suitable next hop.

For example, a gateway can translate messages between a legacy industrial system and another network service whose protocols differ. In ordinary IP configuration, the default gateway is usually the router used to reach destinations outside the local network. This common usage does not imply that it converts every application's protocol.

The words therefore overlap in some contexts. Explain the actual function: forwarding by network address is routing; translating otherwise incompatible communication conventions is protocol conversion. A switch forwarding within a local network is not automatically an Internet router merely because data passes through it.

Packet switching and routing decisions

Packet switching shares links by dividing communication into manageable units that routers forward independently.

A message is split into packets rather than reserving one end-to-end circuit for its entire duration. A router receives a packet, inspects its destination address and forwards it according to its routing/forwarding table. Packets from different users can share the same link at different times. Packets belonging to one exchange need not all take the same route, and can be delayed, lost or arrive out of order. Higher-layer protocols can provide ordering and retransmission where needed; IP forwarding alone does not guarantee them.

At this level, distinguish control information from payload:

Different protocols place fields in different headers. In particular, a TCP sequence number belongs to TCP, not to every IP header. An IP packet does not universally have an error-check trailer; link frames commonly do. A simplified packet drawing should not turn every useful field into a universal IP field.

Routing tables contain destination-network entries and appropriate next hops/outgoing interfaces. Directly connected destinations can be reached directly; a default route handles otherwise unmatched destinations. Routers can learn paths from neighbouring routers using routing protocols and update their information as reachability changes. Routing policies and metrics can influence the route; the geographically shortest path is not guaranteed.

For example, in a simplified table, a destination in network N goes through next hop R1 while all other remote destinations use R2. The router makes that local decision. R1 independently decides its own next hop. The destination host, rather than every router, finally processes the application data. Each hop takes time, and congestion can queue packets before forwarding.

URLs, domain names and FQDNs

Human-readable names identify resources conveniently, while network addresses support delivery.

An IP address is a logical numerical address used to identify a network interface for IP communication. It is not a permanent statement of a person's identity. A domain name is a name in the hierarchical domain-name system, such as example.org. A fully qualified domain name (FQDN) specifies a name completely through the hierarchy to the root, such as library.example.org. The final root dot can be written explicitly as library.example.org.; user interfaces commonly omit it.

Read the hierarchy from right to left. In library.example.org, org is the top-level domain, example is below org, and library is below example.org. Administration can be delegated at different levels, and another subdomain such as revision.example.org can be managed under the same parent. The dots separate labels, not IP octets.

A uniform resource locator (URL) identifies how and where to access a resource. In https://library.example.org/revision/networking.html, https specifies the access scheme, library.example.org is the host name, and /revision/networking.html is the path. A URL may also specify a port, query or fragment. The host can be a domain name or, in some URLs, an IP address.

The whole URL is not an FQDN. DNS can look up the host name; it does not translate the path into an IP address or fetch the page. The application uses the path in its request after reaching the appropriate host. Separating these jobs explains why a correctly resolved address can still return “resource not found”.

DNS name resolution

The Domain Name System provides a distributed naming service, including finding IP addresses associated with host names.

Keeping one complete, continually updated list on every computer would be difficult at Internet scale. DNS distributes responsibility through the hierarchy. Authoritative name servers hold records for the parts of the namespace they serve. A resolver obtains answers for a client and can cache them for the record's permitted lifetime.

For an uncached lookup of library.example.org, a recursive resolver can:

1. Consult a root name server to find the servers responsible for org.
2. Consult an org server to find the authoritative servers for example.org.
3. Ask the appropriate authoritative server for the requested address record.
4. Return the address to the client and cache the result as allowed.

These intermediate answers are often referrals to other name servers, not the final host address. The client normally asks its configured resolver rather than personally performing every step. Cached records can bypass some or all of this chain, reducing delay and DNS traffic. Records expire, so changes may not appear in every cache immediately.

The client can then send IP traffic to the returned address. DNS servers do not route every data packet or carry the website's contents merely because they resolved its name. A name can be associated with several addresses, and an address can serve several names; do not assume a universal one-to-one mapping.

Internet registries and coordinated identifiers

Internet identifiers need coordinated administration so independent networks do not accidentally claim the same globally unique resources.

For number resources, global administration allocates large blocks of IP addresses to regional Internet registries. Regional registries allocate or assign resources onward under their policies, including to network operators and providers. The allocation hierarchy helps maintain unique public-address assignments and records of who is responsible for them. Routers use network reachability information to deliver packets; an allocation record is not itself a live routing decision.

Domain-name registries administer names within particular domains, such as a top-level domain. Registrars provide registration services through which applicants register names. Coordinating the namespace prevents two unrelated registrants independently owning the same exact name under the same parent, and maintains delegation information so DNS can find the responsible servers.

These are administrative services, not one central database that resolves every lookup. The global DNS root is coordinated, but authority is delegated. Number registries coordinate address resources, while domain registries coordinate names. An organisation can change the host address behind its registered domain by updating DNS records without registering a new name for every change.

Without coordinated allocation, duplicate public addresses could make delivery ambiguous. Without coordinated names and delegation, clients could not reliably discover which service is responsible for a particular name. Private addresses used within separate local networks are a different case and are not claimed as globally unique public assignments.