How the Internet Works

Find out what happens when data crosses the internet. Follow a message as it is split into packets, addressed with IP and forwarded from router to router, and see how DNS turns names into addresses. Learn how layered protocols such as TCP and UDP deliver data, how the web, email and streaming run on top of them, how HTTPS protects a connection, and who coordinates and governs a network that no one owns. Worked examples with real numbers show how to calculate packet counts, transfer times and delays.

What you will learn

1Describe the internet and how data crosses it in packets
Describe the internet as a network of networks, explain why data travels in small packets that share links, and measure how fast and how quickly data arrives.
- Identify the parts of the internet
  Describe the devices at the edge of the internet, the links and routers in its core, and the providers and cables that connect them.
- Split messages into packets
  Explain what a packet contains, why data is sent in many small packets, and how the receiver rebuilds the original message.
- Compare packet switching with circuit switching
  Contrast reserving capacity along a path in advance with sharing links packet by packet, and explain why the internet uses packet switching.
- Measure speed and delay on a network
  Use bandwidth, latency and throughput correctly, convert between bits and bytes, and calculate simple transfer and delay times.
2Find computers on the internet by address and name
Read and write IP addresses, follow how routers forward packets toward their destination, and explain how the Domain Name System turns names into addresses.
- Read and write IP addresses
  Explain what an IP address identifies, write IPv4 and IPv6 addresses in binary and in shorthand, and tell public addresses from private ones.
- Follow a packet from router to router
  Describe how routers forward each packet hop by hop using forwarding tables, and how routing protocols keep those tables up to date.
- Look up a domain name with DNS
  Explain how the Domain Name System turns names such as www.example.com into IP addresses through a hierarchy of servers and caches.
3Explain how layered protocols deliver data reliably
Explain what a protocol is, how the internet's protocols are organized in layers, and how TCP and UDP carry data between programs on top of best-effort IP.
- Explain protocols, layers and encapsulation
  Define a protocol, describe the job of each layer of the internet's protocol stack, and follow how each layer wraps the data in its own header.
- Deliver data reliably with TCP
  Explain how TCP uses ports, connections, sequence numbers, acknowledgments and retransmission to turn best-effort packets into a complete, ordered stream of bytes.
- Choose between TCP and UDP
  Compare TCP with UDP and match each to the applications whose needs it suits.
4Explain how the web and other applications run on the internet
Explain the web as one application of the internet, follow an HTTP exchange and a complete secure page load, and describe how email, messaging, peer-to-peer and streaming applications use the network.
- Tell the web apart from the internet
  Explain the web as an application built from URLs, HTTP and linked documents, and describe how clients, servers and content delivery networks serve it.
- Follow an HTTP request and response
  Read the parts of HTTP requests and responses, interpret status codes, and explain how cookies give memory to a stateless protocol.
- Trace a secure page load from URL to screen
  Put DNS, TCP, TLS and HTTP together in order to explain what happens after a URL is entered, and what HTTPS does and does not protect.
- Send email and other messages
  Follow an email from sender to recipient with SMTP, IMAP and MX records, and compare client-server messaging with peer-to-peer and streaming applications.
5Explain who runs the internet and how it is governed
Describe how the internet is coordinated without a single owner, what the main organizations do, and the main policy debates about how it should be run.
- Identify who coordinates names, numbers and standards
  Describe what must be coordinated on the internet and what ICANN, IANA, the regional internet registries, the standards bodies, network operators and governments each do.
- Discuss the main internet governance debates
  Explain the main policy questions about how the internet should be run, and weigh the arguments on each side.

How the Internet Works study guide

5 chapters

Describe the internet and how data crosses it in packets

Data crosses the internet through equipment owned by many different organizations. Follow the links from a host into provider networks, then examine how small packets share those links. The way capacity is shared explains queues and losses. Calculating packet counts, transfer times, and signal travel times helps separate a high advertised rate from a quick response and shows why different applications need different network behavior.

1.1 Identify the parts of the internet

The internet joins independently operated networks through shared protocols, especially the Internet Protocol (IP). No single organization runs the whole system.

Devices and links

A host, or end system, runs applications. Phones, laptops, servers, smart TVs, and sensors can be hosts. A client requests a service; a server provides it. One computer can request a file while supplying another file to someone else.

Links carry bits through copper cable as electrical signals, through optical fiber as light, or through wireless connections as radio. One route can cross several media.

Device	Main job
Switch	Moves frames within a local network
Router	Moves IP packets between networks

A home router box usually combines routing, switching, and a Wi-Fi access point, sometimes with a modem. These remain different jobs even when one box performs them.

Connect to a provider

An access network connects a home, organization, or phone to its provider. Examples include fiber, cable, DSL, office Ethernet, and mobile networks. An internet service provider (ISP) carries traffic between access networks and other networks.

Providers exchange traffic directly at an internet exchange point (IXP), a shared facility that can span several buildings. They can also buy transit, paying another provider to carry traffic onward. Large content companies operate their own networks and data centers to bring services closer to users.

Most intercontinental data crosses submarine fiber-optic cables on the seabed; satellites carry a smaller share. A wireless access link can therefore lead into a long, mainly wired route.

Follow a video call

A café phone calls someone on a mobile network overseas:

Wi-Fi carries packets to the café's access point.
The café router sends them over its access connection to the ISP.
Provider networks carry them onward, perhaps exchanging them at an IXP.
A submarine cable can carry the ocean crossing.
The destination provider delivers them through its mobile network to the other phone.

Different organizations operate these stages. Compatible protocols let them pass packets onward, while each operator manages its own part. The precise path can change as networks alter their connections or respond to failures.

Key Takeaways

Hosts run applications; clients request services and servers provide them.
Links carry bits over copper, fiber, or radio.
Switches connect devices locally; routers connect networks.
ISPs exchange traffic directly or buy transit.
Wireless access often leads into a route carried mainly by fiber.

1.2 Split messages into packets

Large files travel as many packets. Each packet carries a piece of data, allowing other conversations to take turns on the same links.

Separate header and payload

A packet's header holds control information; its payload holds data. An IP header includes source and destination addresses. Routers use the destination to select a next link without interpreting the photo or message inside. The receiving application interprets the delivered bits.

The header resembles an envelope's address, while the payload resembles its contents. Dividing a file into small packets also limits how much must be resent after a loss. A common path limit is a 1,500-byte IP packet, derived from standard Ethernet. Some paths support larger sizes.

Count packets and overhead

A photo contains 3,000,000 bytes. Each packet holds up to 1,460 photo bytes plus 40 header bytes. Divide by payload capacity:

3,000,000 \div 1,460 \approx 2,054.8

Round up to 2,055 packets because a remainder still needs delivery. After 2,054 full payloads, the last contains:

3,000,000 - 2,054 \times 1,460 = 1,160 bytes

Every packet needs a header, including the last:

2,055 \times 40 = 82,200 bytes

Together, data and headers total 3,082,200 bytes. Headers add about 2.7% to the photo size. Dividing by 1,500 would undercount packets because 40 bytes of each packet are unavailable for photo data. Counting the last payload as full would instead invent data that the photo does not contain.

Rebuild the byte order

Packets usually follow the same route, but IP allows different routes. Packets can arrive out of order, arrive twice, or be lost, including when a router's buffer fills.

Transmission Control Protocol (TCP) repairs these effects at the end hosts. It numbers bytes, allowing the receiver to identify positions and gaps. The sender retransmits missing data, and the receiver delivers bytes in order. Receiving the final piece does not fill an earlier gap.

IP only moves individual packets; it does not rebuild the message or perform that recovery. User Datagram Protocol (UDP) also omits retransmission and reordering. Applications using UDP add any recovery they need.

Key Takeaways

A header carries delivery information; a payload carries data.
Divide by payload capacity and round up to count packets.
Each packet adds its own header overhead.
IP packets may be lost, duplicated, or reordered.
TCP restores byte order and retransmits missing data at the endpoints.
UDP applications add recovery themselves when they need it.

1.3 Compare packet switching with circuit switching

Link capacity can be reserved for a connection or shared among arriving packets. The difference matters when traffic pauses and when several users send together.

Reserve or share

Circuit switching reserves a fixed share along the entire path. Traditional landline calls kept that share even during silence.

Packet switching shares capacity on demand, called statistical multiplexing. Each packet uses the link's rate during its turn. This suits bursty traffic: web browsing and messaging send short bursts separated by pauses.

Feature	Circuit switching	Packet switching
Capacity	Reserved per connection	Shared among arriving packets
Silence	Reserved share stays idle	Other traffic can use capacity
Excess demand	New connections may be refused	Queues grow and packets may be lost
Delay	Predictable after setup	Varies with traffic

Store packets while a link is busy

A router uses store-and-forward: it receives a whole packet before sending it onward. Packets wait in a queue when the outgoing link is busy. A memory buffer holds that queue; when full, it cannot accept new arrivals, which are dropped.

A 1 Mbit/s link can reserve ten shares of 100 kbit/s:

1,000 kbit/s \div 100 kbit/s = 10

With packet sharing, 35 users can share the link if each is active only 10% of the time. For independently active users, the probability that more than ten are active together is about 0.0004. Independence matters: synchronized bursts behave differently. Brief queues can form below ten active users too, when packets arrive together.

Where packet switching began

Paul Baran at RAND and Donald Davies at the UK's National Physical Laboratory developed packet-switching ideas independently in the 1960s. Davies coined "packet"; Leonard Kleinrock developed early mathematical work on network queues.

On October 29, 1969, ARPANET sent its first message from UCLA to the Stanford Research Institute. It crashed after "LO" of "LOGIN." ARPANET aimed to share research computers. Baran's separate work studied resilient communications under attack, helping explain the later nuclear-war story.

Packet networks can use alternative routes after failures when another path is available. Today voice over IP carries calls as packets, including calls on 4G and 5G networks.

Key Takeaways

Circuit switching reserves capacity, including during silence.
Packet switching shares links on demand through statistical multiplexing.
Store-and-forward receives a packet before sending it onward.
Busy links create queues; full buffers cause packet loss.

1.4 Measure speed and delay on a network

A connection's speed has more than one meaning. A large transfer needs a high delivery rate; a conversation needs short, steady delays. Keeping bandwidth, throughput, and latency separate helps explain both.

Capacity and actual delivery

Bandwidth is a link's capacity in bits per second. SI prefixes use powers of ten: 1 kbit/s is 1,000 bit/s; 1 Mbit/s is 1,000,000 bit/s; 1 Gbit/s is 1,000,000,000 bit/s. "Mbps" is another common spelling of Mbit/s.

Throughput is the rate actually delivered. The slowest link on a path is its bottleneck. A server link of 1 Gbit/s and a provider link of 10 Gbit/s cannot push a download through a 50 Mbit/s home line faster than 50 Mbit/s. Sharing capacity and protocol overhead reduce the useful rate further.

File sizes use bytes, written with uppercase B; a byte contains 8 bits. A 500 MB file contains 4,000 Mbit. At 100 Mbit/s its ideal transfer time is:

500 \times 8 \div 100 = 40 s

The answer of 5 seconds comes from dividing bytes by a rate in bits without converting units.

Four parts of delay

Latency is travel time from one end to the other. Round-trip time (RTT) includes the outward and return journeys. A ping measurement reports a round trip, usually in milliseconds.

At each hop, distinguish four delays:

Processing: reading a header and deciding where to send the packet.
Queuing: waiting for an outgoing link.
Transmission: placing all the packet's bits onto that link.
Propagation: the signal traveling along the link.

Transmission delay depends on packet size and rate:

transmission time = \frac{packet size in bits}{link rate in bit/s}

A 1,500-byte packet has 12,000 bits. It takes 0.12 ms to transmit at 100 Mbit/s, or 1.2 ms at 10 Mbit/s. Across three 10 Mbit/s links, store-and-forward requires at least 3.6 ms of transmission time for that packet, before adding the other delays.

Propagation delay depends on distance and signal speed:

propagation time = \frac{distance}{signal speed}

In fiber, the signal travels at about 200,000 km/s. Crossing 6,000 km takes 30 ms one way and at least 60 ms round trip. Increasing bandwidth cannot shorten the cable.

Match the measure to the application

A download benefits mainly from throughput. Calls and fast games need low latency and low jitter, the variation in delay. Queues can grow at busy times even when a link's advertised capacity stays the same.

A web page may require several exchanges before it is complete. The browser can request many files together, but some requests depend on earlier responses. RTT matters to those dependencies; server work, available throughput, and browser processing matter too. A fast bulk download and a slow-feeling interactive page can therefore occur on the same connection.

Key Takeaways

Use 8 bits per byte and keep bit/s separate from B.
Throughput is limited by the bottleneck and by shared traffic.
Transmission time is packet bits divided by link rate.
Propagation time is distance divided by signal speed.
RTT measures an outward and return trip.
Processing and queues add delay at routers.
Calls need low latency and jitter; large downloads need throughput.

Try chapter 1:Walkthrough 13 Practice Questions 27 Flashcards

Find computers on the internet by address and name

Addresses let packets reach network connections, while names help people find services. Read IPv4 and IPv6 notation, see how private devices share public addresses, and follow a router’s next-hop decision. Then trace a name lookup through DNS. These steps work together: DNS supplies an address, and routers use that address to move packets toward the service, with caches and routing tables keeping the work manageable.

2.1 Read and write IP addresses

An IP address identifies a network interface: a device's connection to an IP network. A laptop can have different addresses on different networks. The address tells the network where to deliver packets, not who a person is.

Read IPv4 and IPv6

IPv4 uses 32 bits, written as four decimal numbers separated by dots. Each number represents 8 bits and ranges from 0 to 255. For example, 192.168.1.10 is 11000000.10101000.00000001.00001010 in binary. A final number of 256 would need more than 8 bits and is invalid.

There are 4,294,967,296 possible 32-bit values, about 4.3 billion. The central unassigned IPv4 pool was exhausted on February 3, 2011. Existing addresses still work; exhaustion means the original central supply of new blocks ended.

IPv6 uses 128 bits, allowing about 3.4 × 10³⁸ values. It has eight groups of four hexadecimal digits. Leading zeros within a group may be omitted, and one run of all-zero groups may be replaced by ::.

2001:0db8:0000:0000:0000:ff00:0042:8329 becomes 2001:db8::ff00:42:8329. Count five visible groups: the double colon replaces three groups to restore eight. Use :: only once so its expansion stays clear. Many hosts run IPv4 and IPv6 side by side.

Understand a prefix

In 203.0.113.0/24, the slash says that the first 24 bits form the network prefix. Eight bits remain:

32 - 24 = 8, 2^{8} = 256

The block spans 203.0.113.0 through 203.0.113.255. An ordinary IPv4 network of this size reserves the first and last values, leaving 254 device addresses. The slash is a count of network bits, not a device number.

Reuse addresses privately

Three IPv4 ranges are reserved for private networks:

10.0.0.0/8
172.16.0.0/12, from 172.16.0.0 through 172.31.255.255
192.168.0.0/16

These ranges are reused in separate homes and organizations and are not routed on the public internet. For example, 172.20.5.4 is private, while 172.32.0.1 falls outside the private 172 range.

Network address translation (NAT) lets private devices share a public IPv4 address. The router rewrites outgoing source addresses and ports and records the translations. It uses the table to send replies to the correct private device. A new connection from outside needs a way through this mapping; the private address alone cannot direct it there.

Join a network automatically

Dynamic Host Configuration Protocol (DHCP) supplies an address lease, a default gateway, and a DNS resolver. The default gateway is the router used to reach other networks. Leases and provider assignments can change, and many devices can share a public address through NAT.

A hardware MAC address serves a different purpose: common Ethernet and Wi-Fi MAC addresses are 48-bit identifiers used for local-link delivery. IP addresses support delivery across networks. An IP address can suggest an approximate location, but shared and changing assignments make it a poor identifier of an individual.

Key Takeaways

IPv4 has 32 bits; IPv6 has 128 bits.
An IPv6 address can use :: once to replace zero groups.
A prefix length counts network bits.
Private IPv4 ranges can be reused on separate private networks.
NAT maps private traffic to a public address and returns replies.
DHCP supplies an address lease, gateway, and DNS resolver.

2.2 Follow a packet from router to router

A packet moves toward its destination one hop at a time. Each router makes its own next-link decision. The sender does not supply a complete itinerary for an ordinary IP packet.

Read a forwarding table

A forwarding table maps destination prefixes to outgoing links. Consider these entries:

Destination prefix	Outgoing link
`203.0.113.0/24`	Link 2
`0.0.0.0/0`	Link 1

A packet addressed to 203.0.113.57 matches the /24 entry and leaves on link 2. A packet to 198.51.100.7 uses link 1. The /0 entry is the default route, used when no more specific entry matches.

When several entries match, choose the longest matching prefix, meaning the most network bits specified. This is about specificity, not the largest printed address or shortest physical distance. Each next router repeats the lookup using its own table.

A host begins an off-network journey by sending its packet to the default gateway. In a home, that is normally the home router. The packet's destination IP address still names the remote endpoint; the local gateway is only its first hop.

Forwarding and routing do different work

Forwarding is the local action of moving a packet to an outgoing link. Routing finds paths and fills the tables that forwarding uses. Routing protocols exchange reachability information and react when links appear or fail. A change may take time to spread, so a usable alternative path is not the same as an immediate, loss-free repair.

An autonomous system (AS) is a network or collection of networks run under one organization's routing policy, identified by an AS number. Providers, universities, and content companies can operate ASes. Inside an AS, the operator chooses its routing methods. Between ASes, the Border Gateway Protocol (BGP) exchanges routes.

BGP decisions reflect business agreements and routing policies as well as path information. A route that crosses fewer kilometers is therefore not guaranteed to be chosen.

Stop loops and inspect hops

IPv4's time to live (TTL) field is a hop limit. Every router reduces it by one. When it reaches zero, the router drops the packet and usually returns an error. This prevents a routing loop from keeping a packet circulating forever.

If a packet starts at TTL 64 and arrives at TTL 52, twelve routers have reduced it:

64 - 52 = 12

Traceroute, called tracert on Windows, sends probes with increasing hop limits. Errors from successive routers reveal responding hops and their round-trip times. A jump from 15 ms to 90 ms may suggest a long link, but timings alone do not identify a submarine cable. Some routers do not answer probes.

IP remains best effort through all these steps. Its routers attempt delivery without guaranteeing arrival, ordering, or uniqueness. End hosts use transport protocols such as TCP when an application needs a complete ordered stream.

Key Takeaways

Forwarding selects a packet’s next link; routing builds the tables.
Routers select the longest matching destination prefix.
The default gateway handles a host’s off-network packets.
BGP exchanges routes between autonomous systems.
TTL limits router hops and prevents endless loops.

2.3 Look up a domain name with DNS

People type names such as www.example.com; routers need IP addresses. The Domain Name System (DNS) supplies the mapping and other records. A service can keep its name while its address changes.

Follow the name hierarchy

Read a domain name from right to left. An unwritten final dot represents the root. Next comes a top-level domain (TLD) such as com, org, jp, or br, then a second-level domain such as example, then optional subdomains such as www or mail.

No single server stores all domain records. Root servers direct resolvers toward TLD servers. TLD servers direct them toward a domain's authoritative servers, which hold that domain's records. The root service has 13 named identities, operated by 12 organizations; in October 2026 these identities were served from more than 2,000 copies worldwide. Thirteen names therefore does not mean thirteen physical machines.

Resolve an uncached name

A laptop asks its recursive resolver for www.example.com. The resolver may belong to an ISP, an organization, or a public DNS service. With no useful cached information, it works through the hierarchy:

Ask a root server, which refers it to the com servers.
Ask a com server, which refers it to example.com's authoritative servers.
Ask an authoritative server, which returns an address such as 192.0.2.10.
Return that answer to the laptop.

The laptop can now connect to the web server. The DNS reply does not contain the web page.

Cache answers for a limited time

Resolvers and devices keep answers in a cache. A record's time to live (TTL) sets how long a cached answer may be reused. A TTL of 3,600 seconds permits one hour of reuse. Most lookups can therefore avoid a fresh trip through the hierarchy.

If an address record changes, an existing cache can still serve its old answer until that entry expires. This delayed visibility is often called "DNS propagation." It does not mean the authoritative server is slowly sending the change to every computer. DNS TTL is a duration in seconds; IP TTL limits hops.

Choose the record type

Record	Meaning
A	IPv4 address
AAAA	IPv6 address
CNAME	Alias pointing to another name
MX	Servers accepting a domain's email
NS	Authoritative name servers

One name can return several addresses, and several names can share an address. An MX lookup finds mail servers, not the recipient's personal device.

Classic DNS commonly uses UDP port 53 for short exchanges and also supports TCP, including for larger answers. Classic queries are unencrypted. DNS over HTTPS protects the query in transit to its resolver; the resolver still processes it. DNSSEC checks authenticity of signed DNS data. Authenticity and confidentiality solve different problems: a signed answer can still be readable on the path.

Key Takeaways

DNS maps names to addresses and other records.
An uncached resolver follows root, TLD, and authoritative referrals.
DNS TTL controls cache lifetime in seconds.
A and AAAA store addresses; MX directs email.
DNSSEC authenticates DNS data; encrypted DNS protects queries in transit.

Try chapter 2:Walkthrough 9 Practice Questions 24 Flashcards

Explain how layered protocols deliver data reliably

Applications need services that individual IP packets cannot guarantee. Protocol layers divide that work, from physical signals to application exchanges. Follow the wrappers added to data, then use TCP byte numbers to find a missing piece and recover it. Comparing TCP, UDP, and QUIC shows how a file download and a live conversation can use the same internet while making different choices about completeness and delay.

3.1 Explain protocols, layers and encapsulation

A protocol specifies message formats, message order, and what participants do when they send or receive messages. The internet divides these jobs among layers, so one application can work over many kinds of network.

Give each layer a job

The five-layer description separates application needs from the physical signals carrying bits.

Layer	Job	Examples
Application	Defines exchanges for a service	HTTP, DNS, SMTP
Transport	Carries data between programs	TCP, UDP
Network	Delivers packets across networks	IP
Link	Delivers frames over one local link	Ethernet, Wi-Fi
Physical	Transmits signals representing bits	Electrical, light, radio

Hypertext Transfer Protocol (HTTP) defines web requests and responses. It can use TCP without needing to know whether the next link is Wi-Fi or fiber. Replacing that link can leave the web application unchanged. The four-layer TCP/IP description groups link and physical together. The OSI reference model uses seven layers; it is another way of describing networking responsibilities.

Wrap data for delivery

Encapsulation adds information as data moves down the stack. TCP treats application bytes as a stream and divides them into segments, each with a TCP header. IP wraps a segment in an IP packet. Ethernet or Wi-Fi wraps the packet in a frame for one link. Ethernet also adds an error-checking trailer.

At the destination, decapsulation reverses the wrapping. Each layer reads its own information and passes the remaining contents upward. A higher layer communicates with its counterpart through these lower-layer services; the layers are not a line of separate machines between sender and receiver.

An HTTP message can span several TCP segments, and a segment can include bytes from more than one message. TCP's boundaries come from transporting a byte stream, not from the application's message boundaries.

Count the added bytes

Suppose a segment carries 1,460 bytes of web data. A 20-byte TCP header and a 20-byte IPv4 header give:

1,460 + 20 + 20 = 1,500 bytes

Ethernet adds a 14-byte header and 4-byte trailer, making a 1,518-byte frame. Its 58 bytes of headers and trailer are about 3.8% of the frame. These counts describe the stated headers; optional fields can make TCP and IP headers longer.

What changes at a router?

A host runs the whole stack. A router normally handles physical, link, and network functions. It removes the incoming frame, reads the IP destination, chooses an outgoing link, and creates a new frame for that link. A local switch normally works at the link layer.

Frames change hop by hop. End-to-end IP addresses stay the same along an ordinary route, except where NAT rewrites them; other IP fields, including the hop limit, can change. TCP processing belongs to the endpoints.

Many applications above and many link technologies below share IP. This is the internet's narrow waist: a new application need not redesign the core network. The Internet Engineering Task Force (IETF) publishes open protocol documents called Requests for Comments (RFCs). TCP's specification is RFC 9293, published in 2022.

Key Takeaways

Layers divide networking into application, transport, network, link, and physical jobs.
Encapsulation wraps data; decapsulation removes the wrappers.
TCP segments carry stream bytes, not necessarily whole application messages.
Routers rebuild link frames at each hop.
Shared IP lets many applications work over many link technologies.

3.2 Deliver data reliably with TCP

A file needs all its bytes in the right order even when the network drops or reorders packets. Transmission Control Protocol (TCP) supplies that ordered byte stream using cooperation between the two end hosts.

Identify the programs and open a connection

An IP address identifies a network interface. A port identifies a service or communication endpoint on that host. Ports are 16-bit numbers from 0 to 65535. HTTP commonly uses port 80, HTTPS 443, and DNS 53. A client normally chooses a temporary source port.

A TCP connection is distinguished by both IP addresses and both ports. For example, 198.51.100.23:51514 can connect to 203.0.113.80:443 while another connection uses a different client port. This lets a laptop maintain many conversations at once.

TCP opens with a three-way handshake:

The client sends SYN, starting its sequence numbering.
The server returns SYN-ACK, acknowledging that start and supplying its own.
The client sends ACK.

Ordinary setup adds about one RTT before data can begin. Later, each endpoint can close its sending direction with FIN.

Number bytes and find a gap

TCP sequence numbers count bytes, not packets. Use numbering from 1 to make an example easy to read; actual connections begin with chosen initial sequence numbers.

Segment	Bytes carried
First	1–1,000
Second	1,001–2,000
Third	2,001–3,000

If all three arrive, acknowledgment 3,001 means every byte through 3,000 has arrived and 3,001 is next.

If the second segment is missing, the receiver keeps acknowledging 1,001 even after the third arrives. It can hold the later bytes but cannot yet deliver them as a continuous stream. A timeout or repeated duplicate acknowledgments leads the sender to retransmit missing bytes. Once the gap is filled, the receiver can advance its acknowledgment and deliver the data in order.

A checksum detects most damaged segments. Corrupt data is discarded and recovered through retransmission in the same way as missing data.

Protect the receiver and the network

Two controls limit how much data a sender has outstanding:

Control	What it protects	How it helps
Flow control	Receiver's memory	Receiver advertises available buffer space
Congestion control	Network capacity	Sender reduces its rate when congestion appears

The receive window prevents a fast sender from overwhelming a slow receiver. Congestion control responds to a busy path, including loss and growing delay, then permits gradual growth when conditions improve. A fast receiver can still be reached through a congested network.

Understand the cost

Routers forward IP packets; they do not repair each TCP stream. Recovery at the endpoints means packet loss can occur while the application still receives correct bytes. Waiting for a missing segment can delay later bytes. If the path remains broken, TCP eventually reports failure. TCP also supplies no encryption: TLS adds that protection for secure web traffic.

Key Takeaways

A TCP connection uses two addresses and two ports.
The handshake is SYN, SYN-ACK, ACK.
Sequence and acknowledgment numbers refer to byte positions.
A cumulative acknowledgment stops at the first missing byte.
Retransmission repairs losses and most detected corruption.
Flow control protects the receiver; congestion control protects the network.
TCP provides ordering and recovery but does not encrypt data.

3.3 Choose between TCP and UDP

A complete file and live audio have different delivery needs. Compare what waiting for a missing piece would achieve before choosing a transport.

Compare TCP and UDP

Feature	TCP	User Datagram Protocol (UDP)
Data view	Ordered byte stream	Separate datagrams
Setup	Handshake	No transport handshake
Loss recovery	Acknowledgments and retransmission	Application supplies any recovery
Ordering	Bytes delivered in order	No ordering guarantee
Congestion control	Built in	Not supplied by UDP itself

UDP's 8-byte header carries ports, a length, and a checksum. It does not acknowledge, retransmit, or reorder datagrams. "Unreliable" means no delivery guarantee; most datagrams can still arrive successfully on a healthy network.

Ask whether late data is useful

A bank-statement download needs every byte, so retransmission is worth waiting for. TCP also suits email, software updates, and HTTP/1.1 or HTTP/2 transfers. Its cost includes setup and waiting behind a gap in the ordered stream.

In live audio, a piece arriving 200 ms late may miss its playback deadline. Skipping a tiny gap can be less disruptive than pausing the conversation. Calls and fast games often use UDP-based transport so their applications can decide what to recover and what to skip.

DNS commonly uses UDP for a short question-and-answer exchange. It is essential infrastructure, so importance is a poor selection rule. What matters is completeness, deadlines, and which protocol handles recovery.

Add reliability above UDP

QUIC runs over UDP and adds reliable streams, congestion control, and integrated encryption. HTTP/3 uses QUIC to deliver web resources reliably. UDP's missing service is supplied by another protocol above it.

QUIC also combines secure setup more efficiently than opening TCP and then negotiating TLS separately. A browser using HTTP/3 therefore has recovery and security services even though its underlying UDP transport does not provide them.

Key Takeaways

Choose transport according to completeness, deadlines, and recovery needs.
UDP sends datagrams without its own setup, retransmission, or ordering.
TCP suits complete files; UDP-based transport can suit live media.
QUIC adds reliable encrypted streams above UDP for HTTP/3.

Try chapter 3:Walkthrough 10 Practice Questions 22 Flashcards

Explain how the web and other applications run on the internet

A browser combines naming, transport, security, and application protocols to display a page. Read a URL and an HTTP exchange, then connect those details into a secure page load. The same network also carries email, chat, calls, file sharing, and streamed video. Each application arranges requests, stored data, and delivery timing in its own way; recognizing those differences helps explain both normal behavior and useful failure clues.

4.1 Tell the web apart from the internet

The internet carries packets between networks. The World Wide Web is one application on it: linked resources requested through HTTP. Email, calls, and games also use the internet.

How the web began

Tim Berners-Lee proposed the web at CERN in March 1989. The first site, info.cern.ch, ran on his computer. CERN released the web software into the public domain on April 30, 1993. The research network that grew into the internet had begun in 1969.

Three ideas work together: Uniform Resource Locators (URLs) name resources; Hypertext Transfer Protocol (HTTP) requests and delivers them; Hypertext Markup Language (HTML) describes documents with hyperlinks. Browsers display HTML, use CSS for appearance, and run JavaScript for interaction.

Read a URL

Consider https://www.example.com:443/docs/guide.html?lang=en#intro.

Part	Meaning
`https`	Scheme: secure HTTP
`www.example.com`	Host name resolved with DNS
`443`	Port; normally omitted because it is the HTTPS default
`/docs/guide.html`	Resource path on the server
`?lang=en`	Query carrying parameters
`#intro`	Fragment handled by the browser

DNS resolves only the host name. The HTTP request carries the path and query; the browser keeps the fragment for local handling, such as locating a section within the page.

Request a service

A browser acts as a client, requesting resources from a server program. One host name can lead to many physical servers supporting a popular site.

Other applications can use HTTP too. A weather app calls a web application programming interface (API) to obtain structured data, often in JavaScript Object Notation (JSON). It presents that data in its own interface instead of displaying HTML.

Reuse nearby copies

A content delivery network (CDN) holds resource copies in several locations. DNS or routing directs users to appropriate servers, often nearby, reducing latency and sharing the load.

A browser cache stores previously downloaded resources on the user's device. Reusing a usable cached image can avoid fetching it. If a fresh download is needed, a nearby CDN copy can shorten the journey. The two caches save work at different places: one on the device, the other on network servers.

Key Takeaways

The web is one application on the internet.
URLs identify resources using a scheme, host, path, and optional parts.
DNS handles the URL’s host name; the browser handles its fragment.
Clients request resources and servers return responses.
CDNs and browser caches reduce repeated or distant transfers.

4.2 Follow an HTTP request and response

HTTP defines an exchange between a client and a server. A request says what the client wants; a response reports the outcome and can carry a resource. A single page usually needs many such exchanges.

Read a request

An HTTP/1.1 request can begin:

GET /docs/guide.html HTTP/1.1
Host: www.example.com

The start line contains the method, path, and version. GET asks for a resource. The Host header selects the site, since several host names can share one server address. Headers use name: value lines to supply extra information; an optional body carries submitted data.

GET is intended to retrieve information without changing server state. POST submits data in a body, such as a form or file. PUT can store or replace a resource, and DELETE requests its removal.

Read a response

A successful response can start with:

HTTP/1.1 200 OK
Content-Type: text/html
Content-Length: 5120

After the headers and a blank line comes the body. Content-Type describes how to interpret it; here it is HTML. Content-Length gives a body length of 5,120 bytes. An image response could instead use image/png, while an API could return JSON.

HTTP/2 and HTTP/3 carry this information in binary frames instead of HTTP/1.1's textual form. Methods, status codes, headers, and bodies retain their meanings.

Interpret the outcome

Status class	Meaning	Useful examples
1xx	Information	Interim information
2xx	Success	200 OK
3xx	Redirection or cache-related handling	301 Moved Permanently; 304 Not Modified
4xx	Client-side request problem	403 Forbidden; 404 Not Found
5xx	Server-side problem	500 Internal Server Error; 503 Service Unavailable

A 301 response supplies a new URL in Location. A 304 says a validated cached copy remains current. A 404 means the server answered but did not find a resource at the requested path. A DNS failure or unreachable server produces no HTTP status, because no HTTP response arrives.

Give repeated requests continuity

HTTP is stateless: the protocol does not automatically connect a request to earlier ones. A server can send Set-Cookie with a session identifier. The browser later includes a Cookie header on requests for which that cookie applies. The server uses the identifier to find stored session information, such as a login or cart. Cookies can also support tracking, which browser settings and privacy rules restrict.

Fetching HTML does not fetch an entire website. The browser reads the HTML and requests its referenced images, styles, scripts, and fonts. HTTP/1.1 can reuse a TCP connection for successive requests. HTTP/2 carries multiple exchanges over one TCP connection; HTTP/3 does so over QUIC. A browser's Network tab makes these individual requests, statuses, and timings visible.

Key Takeaways

HTTP requests carry methods; responses carry status codes.
Headers describe an exchange and the body carries its content.
A 404 is a response from a reachable server.
Cookies let applications associate otherwise separate requests.
One displayed page commonly requires many HTTP requests.
HTTP/1.1, HTTP/2, and HTTP/3 share request semantics but encode and transport them differently.

4.3 Trace a secure page load from URL to screen

Entering an HTTPS URL starts a sequence of cooperating protocols. HTTPS means HTTP protected by Transport Layer Security (TLS). To follow the sequence, keep finding a server, connecting to it, protecting the exchange, and requesting content separate.

Follow a fresh connection

For a fresh HTTPS connection using HTTP/1.1 or HTTP/2:

The browser checks caches and, if needed, asks DNS for the host's IP address.
It opens a TCP connection to port 443 with the three-way handshake.
A TLS handshake checks the server and establishes encryption keys.
The browser sends an HTTP GET through the protected connection.
The server returns a response, normally including HTML.
The browser requests referenced resources and draws the page.

All the network exchanges travel in IP packets. HTTP/3 uses QUIC over UDP with integrated TLS, so it does not perform the separate TCP setup in this sequence. Cached answers and reused connections can also skip work.

If DNS, TCP setup, TLS setup, and the first HTTP exchange each take one 60 ms round trip, the total before content starts is about:

4 \times 60 = 240 ms

That example omits server processing and other delays. A higher data rate cannot remove those four stated waits, though avoiding a lookup or reusing a connection can remove an exchange.

What TLS checks

TLS provides encryption, hiding content from observers on the path; integrity, detecting changes made in transit; and authentication, checking the server for the requested name.

A certificate binds a domain name to a public key. A trusted certificate authority (CA) signs it. The browser checks the name, validity period, and trust chain, and the server proves possession of the corresponding private key. A copied certificate alone cannot provide that proof because certificates are public.

A name mismatch, expired certificate, or untrusted signer can produce a browser warning. Do not bypass a certificate warning: doing so removes protection against an impostor server.

Connection protection and site trust

HTTPS protects page contents, URL paths and queries, cookies, passwords, and submitted forms in transit. It still exposes the server's IP address and traffic amount and timing. The domain name also remains visible in many connections, depending on DNS and handshake protection.

A secure-connection indicator says the connection is protected for that domain. A dishonest site can obtain a valid certificate for its own name. Check that the domain is the one you intended; an ordinary certificate verifies domain control, not the honesty of every page. Plain HTTP, commonly on port 80, leaves the application exchange readable and open to alteration along the path.

Locate the stage that failed

"Server not found" can indicate a DNS problem. A timeout can mean packets are not getting through or a server is not answering. A certificate warning points to TLS. A 404 or 500 means an HTTP response arrived: the failure is then about the requested resource or server processing. These clues narrow the problem without treating every failed page as a lost internet connection.

Key Takeaways

A fresh TCP-based HTTPS load uses DNS, TCP, TLS, then HTTP.
TLS encrypts content, checks integrity, and authenticates the server for a name.
Certificate validation also requires proof of the matching private key.
HTTPS protects transit contents while leaving some metadata visible.
An encrypted connection does not establish that a site is honest.
Failure messages can identify DNS, connection, TLS, or HTTP stages.

4.4 Send email and other messages

Applications share the packet network but organize their exchanges differently. Email can wait on servers for an offline reader. A call needs timely audio. A video player can build a buffer and choose a lower data rate.

Follow an email

Simple Mail Transfer Protocol (SMTP) sends mail from an app to a mail server and between mail servers. Internet Message Access Protocol (IMAP) lets a reader work with mail stored on the server and synchronize several devices. Post Office Protocol version 3 (POP3) typically downloads mail to one device.

When amira@example.org emails ben@example.net:

Amira's app submits the message to her outgoing server using SMTP.
That server asks DNS for example.net's MX records.
It sends the message by SMTP to a receiving server named in those records.
The message waits in Ben's mailbox.
Ben's app accesses it using IMAP.

The ben part names a mailbox; example.net names its domain. Several MX entries with preference values can provide alternative receiving servers. Sender and reader need not be online together. With webmail, the browser uses HTTPS to the provider, while the providers still exchange mail through SMTP.

SMTP's original design did not verify claimed sender identities. Sender Policy Framework (SPF), DomainKeys Identified Mail (DKIM), and Domain-based Message Authentication, Reporting, and Conformance (DMARC) help receiving servers check domain-related authorization and authenticity. Passing these checks does not make every message welcome or honest.

Messaging and peers

Many messaging apps use a provider's servers to relay messages and hold them for offline recipients. With end-to-end encryption, the intended endpoints hold the content keys. A relay server can pass on encrypted content without reading it. This differs from encrypting only the connection to the provider, where the provider can still process plaintext content.

In peer-to-peer (P2P) communication, hosts take both client and server roles. A BitTorrent peer downloads file pieces from other peers while uploading pieces it has. More peers can add upload capacity. Video calls may send media directly between devices, but NAT can make a relay necessary. Using a relay for a call does not turn every other internet application into P2P.

Adapt a video stream

Streaming services offer short video chunks at several quality levels. The player measures throughput and requests a suitable version of each next chunk over HTTP. It can lower quality when available capacity falls, reducing the risk of an empty playback buffer.

At a constant 5 Mbit/s, one hour consumes:

5,000,000 bit/s \times 3,600 s = 18,000,000,000 bits

Divide by 8 to convert to bytes: 2,250,000,000 bytes, or 2.25 GB. This counts the stated stream data; extra protocol traffic can add usage. Multiplying megabits by seconds gives megabits, so a result labeled MB or GB still needs the bit-to-byte conversion.

Key Takeaways

SMTP sends mail; IMAP and POP3 provide ways to read it.
DNS MX records locate a domain’s receiving mail servers.
End-to-end encryption keeps content keys at the intended endpoints.
P2P peers can both request and supply data.
Adaptive streaming changes chunk quality to fit measured throughput.

Try chapter 4:Walkthrough 12 Practice Questions 30 Flashcards

Explain who runs the internet and how it is governed

The internet works through cooperation among independent networks and organizations. Some bodies coordinate identifiers, others write standards, and operators and governments make different kinds of decisions. Distinguishing these roles makes policy debates easier to follow. Examine how equal traffic treatment, concentrated infrastructure, encryption, and affordable access create benefits and costs for several groups at once.

5.1 Identify who coordinates names, numbers and standards

Independent organizations operate the internet's networks. They need common identifiers and shared technical standards to exchange traffic. Governance coordinates these shared resources and the policies affecting their use.

Coordinate identifiers

Public IP address blocks and domain registrations need coordinated assignments. Conflicting public assignments would make delivery ambiguous. Private IP ranges are deliberately reusable because they are kept within separate private networks.

Internet Corporation for Assigned Names and Numbers (ICANN) is a nonprofit founded in 1998 and based in the United States. It coordinates top-level DNS policy, including generic top-level domains and registrar accreditation. Its affiliate Public Technical Identifiers (PTI) performs the Internet Assigned Numbers Authority (IANA) functions: maintaining the DNS root zone, central address and AS-number pools, and protocol-parameter registries such as ports.

US government contractual oversight of the IANA functions ended on October 1, 2016. The stewardship transition placed oversight with the global communities using those functions through the relevant processes.

Distinguish names from numbers

Five regional internet registries (RIRs) distribute IP addresses and autonomous-system numbers to providers and organizations:

RIR	Region
AFRINIC	Africa
APNIC	Asia-Pacific
ARIN	Canada, United States, parts of the Caribbean
LACNIC	Latin America, parts of the Caribbean
RIPE NCC	Europe, Middle East, parts of Central Asia

Domains follow a different chain. A registry operates a TLD; Verisign operates .com. A registrar sells domain registrations, and a registrant holds a registration for a period, commonly renewed yearly. Country-code domains such as .jp and .br have their own operators and national rules. RIRs do not register domain names.

Write standards and run networks

The Internet Engineering Task Force (IETF) develops protocols such as IP, TCP, DNS, HTTP, and TLS through open work published as RFCs. It values rough consensus and working implementations. Anyone can participate.

The World Wide Web Consortium (W3C), founded by Tim Berners-Lee in 1994, develops web standards including CSS and the Web Content Accessibility Guidelines (WCAG). WHATWG maintains HTML. The Internet Society (ISOC) supports the IETF and promotes an open internet.

Operators choose connections, make transit agreements, and announce routes with BGP. Twelve organizations operate the DNS root service. Governments set territorial laws on telecommunications, competition, privacy, and content. The International Telecommunication Union (ITU) coordinates telecommunications work and radio spectrum internationally; it does not register domains or allocate internet address blocks.

Include several kinds of participant

A multistakeholder model involves governments, companies, civil society, the technical community, and academics. The Internet Governance Forum (IGF) provides discussion across these groups; it does not issue binding decisions. First held in 2006, it became a permanent UN forum in December 2025. Coordination through these bodies gives the network shared rules without creating one organization that owns every link or controls every service.

Key Takeaways

ICANN coordinates top-level DNS policy; PTI performs IANA functions.
RIRs distribute address blocks and AS numbers, not domain names.
A registry operates a TLD, a registrar sells registrations, and a registrant holds one.
IETF develops internet protocols; W3C develops web standards; WHATWG maintains HTML.
Operators run networks and governments make territorial law.
Multistakeholder governance includes several sectors; IGF discussions are nonbinding.

5.2 Discuss the main internet governance debates

Internet policy choices affect users, network operators, online services, and governments differently. A useful argument identifies who makes the decision, traces the effects, and weighs a benefit against a cost.

Equal treatment and zero-rating

Net neutrality is the principle that ISPs should treat lawful traffic equally, without favoring particular services through blocking, slowing, or special charges. Supporters connect equal treatment to competition and expression: a new service should not need a favorable deal with every ISP. Opponents argue for flexibility in network management and service offers and warn that strict rules can discourage investment. National rules differ and can change.

Suppose a mobile operator zero-rates one video app: watching it does not count toward a data allowance. Users of that app gain usable data. The favored app gains an advantage over rivals whose traffic still counts. The operator can market the offer and may receive commercial benefits. A regulator must weigh affordability against fair competition. Stating only that "free data is good" misses the effect on competing services; stating only that "the offer is unfair" misses the user's immediate benefit.

Concentration and government control

A few large cloud and content-delivery companies can serve huge numbers of sites efficiently. They can also create shared dependencies: one provider's outage may interrupt many otherwise unrelated services. Concentration is therefore a question about both efficiency and resilience.

Governments can filter sites, demand domestic data storage, or shut down access. Supporters may cite security and public order; critics point to expression, livelihoods, education, and the harm to users unrelated to an incident. Fragmentation, sometimes called the "splinternet," describes the risk of increasingly separated national networks.

Scarce addresses and protected data

The central IPv4 pool is depleted. Transfers of existing blocks and extensive NAT keep IPv4 usable, while IPv6 provides a much larger address space. Moving takes time because equipment and services must continue working together during coexistence.

Encryption protects private information against interception and surveillance. Some law-enforcement agencies seek special access to encrypted data. Many security specialists argue that built-in access also creates weaknesses that others could exploit. Evaluating the proposal requires considering both investigation needs and the security of ordinary users. Data-protection rules vary between countries.

Access is more than coverage

The ITU estimated about 6 billion people online in 2025 and 2.2 billion offline, with much of the gap in lower-income countries and rural areas. Affordability, infrastructure, and usable access shape who benefits from online services. A network reaching a village does not help everyone equally if its price remains out of reach.

Match a proposal to the responsible actor. Registries and ICANN coordinate names; RIRs coordinate addresses; standards bodies develop technical rules; operators choose network connections and services; governments make law. A domain-registration policy, a protocol design, and a national access restriction can interact while remaining different kinds of decision.

Key Takeaways

Policy arguments should identify stakeholders, effects, benefits, and costs.
Zero-rating lowers a user’s counted usage while favoring selected services.
Concentrated infrastructure brings scale and shared outage risks.
Government restrictions raise security, expression, and fragmentation questions.
IPv6, encryption, and affordable access address different governance needs.

Try chapter 5:Walkthrough 6 Practice Questions 16 Flashcards

How the Internet Works examples

5 Walkthroughs

Follow packets and calculate the time they need

18 steps

Trace a network journey, divide a message into packets, and distinguish sharing, capacity, and delay.

1
IntroductionStart with a message and a path
A message passes through independently operated networks between two hosts. Follow the links, divide its data into packets, and calculate the time needed to carry them.

Keep capacity and travel time separate: a route can carry many bits each second yet still take time to cross a long distance.
2
ExplanationGive devices their jobs
A host runs applications. A client requests a service; a server provides it. A router forwards packets between networks, while a switch connects local devices.

One home box can combine routing, switching, and Wi-Fi access. These are distinct jobs even when the same device performs them.
3
ExampleCross more than one network
A café phone sends packets by Wi-Fi to its access point, then through its router and ISP. Provider networks carry them onward, perhaps through an IXP and a submarine cable, to the other person's mobile network.

Radio at the two ends can surround a long, mainly fiber route.
4
ChallengeIdentify the first link
Does a laptop's Wi-Fi connection tell you how its data crosses an ocean? Explain.

---reveal---

No. Wi-Fi describes the local link to an access point. Later links can use different media; most intercontinental traffic crosses submarine fiber cables.
5
AnalogySeparate the envelope from its contents
A packet resembles an addressed envelope: the header helps delivery, and the payload carries data. A router can choose the next link without interpreting the photograph inside.

A large photo needs many packets, each with a header. The receiver places their data in the right positions.
6
ExampleCount full payloads first
A photo has 3,000,000 bytes and each packet can carry 1,460 photo bytes. Divide the data size by the payload capacity:
$3,000,000 \div 1,460 \approx 2,054.8$
The result means 2,054 full payloads and a remainder. A remainder cannot be left behind, so the photo needs 2,055 packets. Use payload capacity for this division; the packet’s header space cannot carry photo bytes.
7
ExampleAccount for the final piece
The last packet’s data is what remains after filling 2,054 payloads:
$3,000,000 - 2,054 \times 1,460 = 1,160 bytes$
If each packet adds 40 header bytes, all 2,055 packets together add 82,200 bytes of headers. The short final payload still has a header. Its unused payload capacity is not extra photo data, so do not count every packet as full.
8
ChallengeApply the packet-count method
A 5,250-byte report uses packets carrying at most 1,400 data bytes each. Every packet adds a 40-byte header. How many packets are needed, and how many header bytes are added?

---reveal---

Three full payloads carry 4,200 bytes, leaving 1,050 bytes. That requires four packets. Four 40-byte headers add 160 bytes. Counting three packets would leave part of the report unsent; header space is separate from the payload capacity.
9
ExplanationShare pauses as well as bursts
Circuit switching reserves a fixed share for a connection, even during silence. Packet switching gives arriving packets turns on the link.

This statistical multiplexing suits bursty web traffic: one user's pause leaves room for another. Overlapping bursts can still create waits.
10
ExplanationWatch a queue form
A router receives a whole packet before forwarding it: store-and-forward. Packets wait in a buffer while the outgoing link is busy.

If arrivals exceed the sending rate, the queue grows. Once the buffer fills, new packets are dropped. Brief waits can also follow simultaneous arrivals.
11
ChallengeExplain a pause during a call
Several users start large transfers over the same packet-switched access link. A call on that link becomes delayed, although the advertised bandwidth has not changed. How can both facts be true?

---reveal---

The link still has the same capacity, but more packets compete for it. Call packets can wait behind other traffic in a queue. If the buffer fills, some packets may also be lost. Shared demand changes delay without requiring a change in the physical link rate.
12
ExplanationName the measurement
Bandwidth is link capacity; throughput is actual delivered rate. The bottleneck limits that rate.

Latency is travel time, RTT is there and back, and jitter is variation in delay. Downloads mainly need throughput; live calls need short, steady delays.
13
ExampleConvert the file before dividing
A 500 MB file crosses a 100 Mbit/s link. First convert bytes to bits using eight bits per byte:
$500 MB \times 8 = 4,000 Mbit$ $4,000 Mbit \div 100 Mbit/s = 40 s$
This is an ideal minimum for the stated data and rate. Overhead and shared traffic can lengthen it. Dividing 500 by 100 to get 5 seconds mixes megabytes with megabits.
14
ExamplePlace one packet onto a link
A 1,500-byte packet contains 12,000 bits. At 10 Mbit/s, its transmission time is:
$12,000 \div 10,000,000 = 0.0012 s = 1.2 ms$
Across three such links with store-and-forward, that one packet needs at least 3.6 ms of transmission time. Each link must send its bits. Processing, queuing, and signal travel add further time, so 3.6 ms is only the transmission contribution.
15
ExampleLet the signal travel
Propagation is travel along a link. For 6,000 km of fiber at about 200,000 km/s:
$6,000 \div 200,000 = 0.03 s = 30 ms$
The return journey makes at least 60 ms of propagation RTT. A higher bit rate can shorten transmission but cannot shorten this unchanged distance. Processing and queuing are separate delays, so a measured RTT can exceed this physical travel minimum.
16
ChallengeChange capacity without changing distance
A link rate doubles while packet size, cable length, and signal speed stay unchanged. What happens to transmission and propagation times?

---reveal---

Transmission time halves because the same number of bits is divided by twice the rate. Propagation time stays the same because distance and signal speed did not change. A faster rate can reduce one part of delay without halving the entire journey.
17
ChallengeCombine packet size and link rate
A 3,000-byte data item uses two 1,500-byte payloads, each with a 40-byte header. Both packets cross one 8 Mbit/s link. Ignoring other traffic and delays, what total transmission time do they need?

---reveal---

The link carries 3,080 bytes including both headers, or 24,640 bits. Dividing by 8,000,000 bit/s gives 0.00308 s, or 3.08 ms. Counting only payload gives 3 ms and leaves out the header bits that also occupy the link.
18
RecapRead a connection by its resources
A host’s application sends data through independently operated networks. Packets share links, wait in queues, and can be lost. Headers occupy capacity along with the payload.

For calculations, first align units: bytes become bits before division by a bit rate. Then ask which time you need—transmission, propagation, or a larger delay including queues and processing. For a call, examine latency and jitter as well as delivered throughput.

Find an address and follow the next hop

17 steps

Read address notation, choose forwarding entries, and trace a DNS answer through the hierarchy and caches.

Save this lesson to follow every Walkthrough step by step.

Save this lesson

1
IntroductionFind a service in two stages
2
ExplanationRead an IPv4 address
3
ExampleShorten an IPv6 address
4
ExampleRead the slash
5
ConnectionJoin privately and reach outward
6
ChallengeCheck a private address
7
ExplanationChoose only the next link
8
ExampleApply two table entries
9
Key InsightBuild tables separately from using them
10
ChallengeUse specificity and a hop limit
11
ExplanationRead the hierarchy from the right
12
ExampleFollow an uncached lookup
13
ExplanationAsk for the right kind of record
14
ChallengeTime an old cached answer
15
ConnectionProtect a lookup in different ways
16
ChallengeSeparate a lookup from delivery
17
RecapKeep names, endpoints, and hops distinct

Wrap data and recover a missing range

16 steps

Follow protocol layers, repair a TCP gap, and choose a transport for the application’s needs.

Save this lesson to follow every Walkthrough step by step.

Save this lesson

1
IntroductionBuild a service above packets
2
ExplanationAssign the five jobs
3
AnalogyWrap one delivery inside another
4
ExampleFollow the bytes through the wrappers
5
ChallengePredict what a router replaces
6
ExplanationIdentify a TCP conversation
7
ExampleLabel three byte ranges
8
ExampleLeave the acknowledgment at the gap
9
ExampleFill the gap and advance
10
ChallengeFind the next missing byte
11
Key InsightProtect two different resources
12
ExplanationDecide whether to wait
13
ChallengeChoose by the data’s useful lifetime
14
ConnectionBuild reliability on UDP
15
ChallengeCorrect a protocol-stack claim
16
RecapLocate the responsibility

Load a secure page and deliver other messages

20 steps

Read an HTTP exchange, assemble a fresh HTTPS load, and compare email, messaging, and streamed video.

Save this lesson to follow every Walkthrough step by step.

Save this lesson

1
IntroductionConnect the services into an application
2
ExplanationSeparate the web from its network
3
ExampleRead the resource address
4
ChallengeChoose where a part is handled
5
ExampleRead a request line
6
ExampleInterpret the server’s response
7
ChallengeDistinguish an error response from no response
8
ExplanationRemember a session through an identifier
9
ExplanationResolve and connect first
10
ExplanationAuthenticate before exchanging content
11
ExampleRequest content inside the protection
12
ExampleCount the stated round trips
13
ChallengeRemove work from a fresh-load model
14
Key InsightRead connection security accurately
15
ExampleHand mail from server to server
16
ExplanationDistinguish a relay from an endpoint
17
ExampleFit video to delivered capacity
18
ChallengeCalculate a shorter stream
19
ChallengeLocate a failure after naming worked
20
RecapFollow the exchange that actually happened

Match governance choices to people and organizations

12 steps

Separate identifier coordination from operations and law, then weigh policy effects on competing stakeholders.

Save this lesson to follow every Walkthrough step by step.

Save this lesson

1
IntroductionAsk who decides what
2
ExplanationCoordinate names and numbers
3
ExampleFollow a domain registration
4
ConnectionSeparate standards, operations, and law
5
ChallengeSend a request to the right place
6
ExampleEvaluate a zero-rated video app
7
ChallengeName a cost as well as a benefit
8
Key InsightNotice a shared dependency
9
ExplanationWeigh encryption access
10
ConnectionConsider who can get online
11
ChallengeCompare two claims about access
12
RecapBuild a reasoned policy answer

How the Internet Works practice questions

50 Practice Questions in 5 sets

Set 1: Recognize the parts and follow simple exchanges10 Practice Questions

Question 1 of 10

A mail server needs the names of servers that accept email for example.org. What DNS record type should it request?

AAAA records
MX records
CNAME records
A records

Show answerHide answer

Answer: B

MX records name a domain’s receiving mail servers. An A record supplies an IPv4 address, but it does not identify which host accepts the domain’s mail.

Question 2 of 10

Why is 198.51.100.260 not a valid dotted-decimal IPv4 address?

IPv4 needs five decimal groups.
IPv4 groups cannot contain zero.
One group exceeds the 8-bit range.
The final group must be a single digit.

Show answerHide answer

Answer: C

Each IPv4 group represents 8 bits, allowing values from 0 to 255. The value 260 is too large; three-digit groups are allowed when their values remain within that range.

Question 3 of 10

A tablet is using a library Wi-Fi network to request a digital journal. What role does the tablet play in this exchange?

A router forwarding traffic
A client requesting a service
A server providing the journal
A switch joining local devices

Show answerHide answer

Answer: B

The tablet is the client because it requests the journal. Using Wi-Fi does not make it a router: routers forward packets between networks, while this tablet is requesting an application service.

Question 4 of 10

A sensor log has 6,100 bytes. Packets carry at most 1,200 log bytes and each adds a 40-byte header. Ignoring other overhead, how many packets and total transmitted bytes are needed?

5 packets; 6,300 bytes
6 packets; 7,440 bytes
5 packets; 6,100 bytes
6 packets; 6,340 bytes

Show answerHide answer

Answer: D

Five full payloads carry 6,000 bytes, leaving 100 bytes for a sixth packet. Six headers add 240 bytes, so the total is 6,340 bytes. Counting only five packets loses the final 100 bytes; treating every packet as full instead overcounts the last payload.

Question 5 of 10

A reserved voice circuit carries no sound for several seconds. What happens to its reserved capacity during that silence?

It remains allocated to the call.
It goes to other users’ packets.
It shrinks to match the sound level.
It is lent out until speech resumes.

Show answerHide answer

Answer: A

A circuit keeps its capacity reserved even during silence. Releasing that share for other users on demand would describe packet sharing, not the behavior of an established reserved circuit.

Question 6 of 10

A laptop changes from Ethernet to Wi-Fi while its web application continues using HTTP and TCP. What feature of layering explains this?

Upper layers can use the same service over different links.
Transport protocols choose a different application for each link.
Wireless links replace the need for network-layer addressing.
Application messages bypass lower layers after a link changes.

Show answerHide answer

Answer: A

Each layer provides a service while hiding the details of how it supplies it. HTTP and TCP can keep their jobs when a different link technology carries their data. Wi-Fi still needs network-layer addressing for internet delivery; replacing a link does not remove IP from the stack.

Question 7 of 10

A TCP receiver has every byte from 1 through 2,400, with no gap. Using this simplified numbering, what cumulative acknowledgment number should it send?

2,399
2,400
2,402
2,401

Show answerHide answer

Answer: D

A TCP acknowledgment names the next byte expected. With bytes 1 through 2,400 complete, that number is 2,401. Sending 2,400 would describe the last received byte instead of the next one; TCP sequence and acknowledgment numbers count byte positions, not segment counts.

Question 8 of 10

For https://portal.example.net/files?sort=date#latest, what part is handled by the browser without being sent as part of the HTTP request target?

/files
?sort=date
#latest
portal.example.net

Show answerHide answer

Answer: C

The fragment, #latest, identifies a part or position for the browser to handle locally. The path /files and query ?sort=date identify the requested server resource and parameters. Confusing a fragment with a query would send the server a responsibility that belongs to the browser.

Question 9 of 10

An HTTP server returns 404 Not Found for a document. What has this response established?

The server answered but did not find that resource.
The browser could not obtain any server address.
The transport connection failed before any reply.
The server redirected the browser to a new path.

Show answerHide answer

Answer: A

A 404 means an HTTP server answered but did not find that resource. A DNS failure can also prevent a page from loading, but it produces no HTTP response by itself.

Question 10 of 10

An organization needs a regional allocation of internet address space. What type of organization handles this part of coordination?

A domain-name registrar
A top-level-domain registry
A regional internet registry
A certificate authority

Show answerHide answer

Answer: C

A regional internet registry distributes address space and AS numbers. Domain registrars handle name registrations instead; the similar names describe separate coordination systems.

Set 2: Apply addresses, rates, transport, and security10 Practice Questions

Save this lesson to see every answer and practice every set.

Save this lesson

Question 1 of 10

A 180 MB archive crosses a path whose bottleneck rate is 60 Mbit/s. Ignoring overhead and other traffic, what is its minimum transfer time?

3 s
6 s
12 s
24 s

Show answerHide answer

Save this lesson to see this answer.

Question 2 of 10

A router has 203.0.113.0/24 → link 2, 203.0.113.128/25 → link 3, and a default route on link 1. A /25 block here spans final values 128–255. Where does 203.0.113.170 go?

Link 2, using the /24 entry
Link 3, using the /25 entry
Link 1, using the default entry
Both links 2 and 3 at once

Show answerHide answer

Save this lesson to see this answer.

Question 3 of 10

A game sends frequent position updates, and an old update becomes useless once a newer one is available. What transport approach fits that timing need?

TCP, waiting for every old update
TCP, with retransmission disabled
UDP, with recovery chosen by the app
UDP, with guaranteed ordered delivery

Show answerHide answer

Save this lesson to see this answer.

Question 4 of 10

A fresh HTTP/2 HTTPS page load has no usable caches or existing connection. In this model, DNS, TCP setup, TLS setup, and the first HTTP exchange each take one 35 ms RTT, with no other delay. What sequence and time fit?

DNS → TCP → TLS → HTTP; 140 ms
DNS → TLS → TCP → HTTP; 140 ms
DNS → TCP → HTTP → TLS; 105 ms
DNS → TCP → TLS → HTTP; 105 ms

Show answerHide answer

Save this lesson to see this answer.

Question 5 of 10

A user wants the same server-held mailbox and read state on a phone and laptop. What mail-reading protocol is designed for this?

SMTP
POP3
IMAP
DNS

Show answerHide answer

Save this lesson to see this answer.

Question 6 of 10

An ISP excludes one music service from data allowances while rival services still count. What competition concern does this raise?

Every service now faces an identical data cost.
The ISP has stopped allocating public addresses.
All music providers must use the same servers.
The favored service gains an advantage over rivals.

Show answerHide answer

Save this lesson to see this answer.

Question 7 of 10

Two independently run provider networks connect at an IXP to exchange traffic directly. What does the IXP provide?

A provider paid to carry traffic onward
A shared facility for traffic exchange
An access network serving one household
A content network serving nearby copies

Show answerHide answer

Save this lesson to see this answer.

Question 8 of 10

A home laptop sends from a private IPv4 address through NAT. A reply reaches the router’s public address and translated port. How does it reach the laptop?

DNS converts the public address into the laptop’s name.
The NAT table maps the reply to the private endpoint.
DHCP changes the reply’s sender to the local gateway.
The remote server broadcasts it to every private host.

Show answerHide answer

Save this lesson to see this answer.

Question 9 of 10

A TCP receiver has bytes 1–800 and 1,601–2,400, but bytes 801–1,600 were lost. What cumulative acknowledgment and delivery behavior fit?

ACK 801; hold the later bytes until the gap is filled.
ACK 2,401; deliver all received bytes as one stream.
ACK 1,601; request the final segment a second time.
ACK 800; discard every byte after the missing segment.

Show answerHide answer

Save this lesson to see this answer.

Question 10 of 10

A 2,000 km fiber path carries a 1,000-byte packet at 8 Mbit/s. Signal speed is 200,000 km/s. Ignoring other delays, what one-way delay includes both transmission and propagation?

1 ms
10 ms
20 ms
11 ms

Show answerHide answer

Save this lesson to see this answer.

Set 3: Connect packet behavior to application needs10 Practice Questions

Save this lesson to see every answer and practice every set.

Save this lesson

Question 1 of 10

A 10,010-byte file is split into payloads of at most 1,000 bytes. How many packets are needed, and how many data bytes are in the last?

10 packets; 1,010 bytes
11 packets; 1,000 bytes
11 packets; 10 bytes
10 packets; 10 bytes

Show answerHide answer

Save this lesson to see this answer.

Question 2 of 10

A packet router’s output runs at 4 Mbit/s. For 2 seconds, input totals 6 Mbit/s. It starts with an empty buffer that can hold 2 Mbit. Ignoring processing, how many bits are dropped during that interval?

2 Mbit
4 Mbit
8 Mbit
12 Mbit

Show answerHide answer

Save this lesson to see this answer.

Question 3 of 10

Two identical 2,400 km fiber routes use 200,000 km/s signals. One route’s link rate doubles, with packet size unchanged and no queues. What happens to propagation and transmission delays?

Both delays are halved.
Propagation halves; transmission stays the same.
Both delays stay the same.
Propagation stays the same; transmission halves.

Show answerHide answer

Save this lesson to see this answer.

Question 4 of 10

An IPv4 packet enters a router with TTL 1. What happens when that router applies the hop-limit rule?

It forwards the packet with TTL 1.
It reduces TTL to zero and drops the packet.
It replaces TTL with a DNS cache lifetime.
It restarts TTL from the sender’s first value.

Show answerHide answer

Save this lesson to see this answer.

Question 5 of 10

A resolver caches an address at 09:00 with TTL 1,800 s. The authoritative address changes at 09:10. No early refresh occurs. At 09:20, what can the resolver return, and when does its original entry expire?

The old address; at 09:30
Only the new address; at 09:30
The old address; at 09:40
Only the new address; at 09:10

Show answerHide answer

Save this lesson to see this answer.

Question 6 of 10

A router receives an Ethernet frame carrying an IP packet and forwards that packet over Wi-Fi, with no NAT. What happens to the link wrapper and destination IP address?

Both the frame and destination IP remain unchanged.
Both the frame and destination IP are replaced.
The frame is replaced; the destination IP remains.
The frame remains; the destination IP is replaced.

Show answerHide answer

Save this lesson to see this answer.

Question 7 of 10

A browser downloads a complete document over HTTP/3 after some underlying UDP datagrams are lost. What explains how reliable delivery can still occur?

IP routers retransmit lost datagrams.
QUIC supplies loss recovery above UDP.
UDP guarantees delivery of every datagram.
HTTP/3 switches to TCP after each loss.

Show answerHide answer

Save this lesson to see this answer.

Question 8 of 10

A browser already has a usable copy of an image stored on the laptop. What mechanism can avoid downloading that image again?

A CDN server cache
A recursive DNS cache
A server session cookie
The browser cache

Show answerHide answer

Save this lesson to see this answer.

Question 9 of 10

A fraudulent store serves HTTPS with a valid certificate for its own domain. What conclusion can a visitor draw from successful certificate validation?

The store’s products and payments have been approved.
The store’s legal owner has been fully investigated.
The connection authenticates that domain’s server.
The network can see every submitted form field.

Show answerHide answer

Save this lesson to see this answer.

Question 10 of 10

Many unrelated services fail during one large cloud provider’s outage. What governance concern does this illustrate?

A failure isolated to one business’s application
Infrastructure shared across many services
A separate routing choice made by each household
A cache expiring independently on each device

Show answerHide answer

Save this lesson to see this answer.

Set 4: Diagnose missing pieces and misleading clues10 Practice Questions

Save this lesson to see every answer and practice every set.

Save this lesson

Question 1 of 10

A frame contains 900 application bytes, a 20-byte TCP header, a 20-byte IPv4 header, and 18 Ethernet bytes. What are the IP packet size and full frame size?

900 bytes; 940 bytes
918 bytes; 958 bytes
958 bytes; 958 bytes
940 bytes; 958 bytes

Show answerHide answer

Save this lesson to see this answer.

Question 2 of 10

A TCP receiver advertises a 64 kB window, but the path becomes congested while the receiver still has room. What should constrain sending, and why?

Only the receive window; space at the host means the path is clear.
Congestion control; receiver space does not measure network capacity.
The receive window must shrink first; congestion control starts only when the host is full.
No further limit; retransmission prevents router buffers from overflowing.

Show answerHide answer

Save this lesson to see this answer.

Question 3 of 10

A server tells a browser that its validated cached copy is still current. What HTTP status reports that outcome?

301 Moved Permanently
404 Not Found
304 Not Modified
503 Service Unavailable

Show answerHide answer

Save this lesson to see this answer.

Question 4 of 10

A person uses a webmail browser to email a different provider’s user, who reads mail with an IMAP app. What sequence describes these three application links?

HTTPS to webmail; SMTP between providers; IMAP to read
SMTP to webmail; HTTPS between providers; IMAP to read
HTTPS to webmail; IMAP between providers; SMTP to read
IMAP to webmail; SMTP between providers; HTTPS to read

Show answerHide answer

Save this lesson to see this answer.

Question 5 of 10

A small publisher registers example.org through a company that sells renewable registrations. What role does the company play, and what role does the publisher play?

Company: registry; publisher: regional registry
Company: registrar; publisher: registrant
Company: registrant; publisher: registrar
Company: regional registry; publisher: registry

Show answerHide answer

Save this lesson to see this answer.

Question 6 of 10

A proposal would give investigators special access to end-to-end encrypted messages by adding a built-in access mechanism. What assessment addresses both the intended benefit and the security concern?

It may aid investigations while introducing a weakness others could exploit.
It can aid investigations without changing who can recover the message plaintext.
It can weaken encryption only when the users’ own endpoint keys are stolen.
It exposes content only to investigators because the access path has one intended purpose.

Show answerHide answer

Save this lesson to see this answer.

Question 7 of 10

A company’s two sites connect through different ISPs. Neither ISP owns the other’s equipment. How can data still cross between the sites?

The providers exchange traffic using shared IP rules.
The providers must operate inside one autonomous system.
The sites must receive settings from one DHCP server.
The providers must reserve an end-to-end circuit for each message.

Show answerHide answer

Save this lesson to see this answer.

Question 8 of 10

A 4,500-byte message uses IP packets of at most 1,500 bytes, each containing a 40-byte header. A student predicts three packets. What result correctly accounts for payload capacity and total header bytes?

3 packets with 120 header bytes
4 packets with 120 header bytes
4 packets with 160 header bytes
3 packets with 160 header bytes

Show answerHide answer

Save this lesson to see this answer.

Question 9 of 10

What is the compressed form of 2001:0db8:0000:0000:0000:0000:002a:0007?

2001:db8::2a::7
2001:db8::2a:7
2001:db8:0:2a:7
2001:db8::2:7

Show answerHide answer

Save this lesson to see this answer.

Question 10 of 10

An IPv4 packet reaches router R with TTL 3. Incorrect forwarding tables send it along R → S → T → R. Each router decrements TTL before sending. Where does the packet expire, and why?

At S, because reaching R already uses another hop before forwarding.
At R on return, because only routers reached after R reduce TTL.
At T after forwarding once more, because zero permits a final hop.
At T before forwarding, because every router reduces the remaining count.

Show answerHide answer

Save this lesson to see this answer.

Set 5: Reason through combined network situations10 Practice Questions

Save this lesson to see every answer and practice every set.

Save this lesson

Question 1 of 10

A 2 Mbit/s link has five users with 500 kbit/s demand each while active. A packet-switched design allows all five to join. What happens when four are continuously active and a fifth begins sending, ignoring protocol overhead?

Each active user retains a reserved 500 kbit/s circuit.
The fifth user’s traffic uses spare capacity at full rate.
The physical link rate grows to match 2.5 Mbit/s demand.
Demand exceeds capacity, so a queue grows and may overflow.

Show answerHide answer

Save this lesson to see this answer.

Question 2 of 10

A 300 MB file crosses links rated 100, 30, and 80 Mbit/s in sequence. With no competing traffic, ignore overhead, startup, and propagation. What rate limit and ideal transfer time apply?

30 Mbit/s and 80 s
100 Mbit/s and 24 s
80 Mbit/s and 30 s
30 Mbit/s and 10 s

Show answerHide answer

Save this lesson to see this answer.

Question 3 of 10

A service moves servers and updates its authoritative A record. A resolver still has an unexpired old record, and the old server no longer runs the service. Why can DNS answer successfully while the page fails, and what can a fresh lookup after expiry change?

The resolver reached the new server; expiry will correct the browser’s local fragment.
The old server must forward the request; expiry will increase its available link rate.
The cache can return an old address; a fresh lookup can obtain the new server address.
The update erased every cached address; expiry will let the resolver resume answering.

Show answerHide answer

Save this lesson to see this answer.

Question 4 of 10

A TCP sender uses segments carrying bytes 1–600, 601–1,200, and 1,201–1,800. The first and third arrive. After the missing middle range is retransmitted successfully, what cumulative ACK can follow without resending the third range?

601
1,801
1,201
1,800

Show answerHide answer

Save this lesson to see this answer.

Question 5 of 10

An application must download an exact complete file and is considering bare UDP. What additional responsibility would the application or another protocol need to take on?

Identify missing file pieces and arrange their retransmission.
Discard out-of-order pieces and assume no missing data remains.
Count received datagrams and accept the file when arrivals stop.
Trust each datagram checksum to recover all absent payload bytes.

Show answerHide answer

Save this lesson to see this answer.

Question 6 of 10

A service puts copies of its images in several regions. Users fetch from closer servers while the original content remains centrally maintained. How can that change performance?

It shortens the distance of every packet in every application.
It removes the need to resolve the service’s host names.
It guarantees that each user receives a higher link rate.
It can reduce latency and distribute requests across servers.

Show answerHide answer

Save this lesson to see this answer.

Question 7 of 10

A browser receives Set-Cookie during login, returns the cookie on a later request, and receives an account page. What explains the remembered login without changing HTTP’s stateless design?

The cookie turns the IP address into a permanent user identity.
Each router stores the user’s earlier HTTP requests.
The server associates the returned identifier with stored session data.
The browser resends the entire password on every page request.

Show answerHide answer

Save this lesson to see this answer.

Question 8 of 10

A captured HTTPS connection shows a server IP address and packet timings but no readable form fields. The certificate matched the requested domain. What does this evidence support?

All metadata and all page content are hidden.
Transit content is protected, but traffic metadata remains visible.
The website’s owner and every form purpose are verified as honest.
TCP alone encrypted the form because its delivery is reliable.

Show answerHide answer

Save this lesson to see this answer.

Question 9 of 10

A video player sees throughput fall below its current encoded stream rate, while lower-quality chunks are available. What response can help avoid an empty playback buffer?

Keep the higher rate and let TCP raise capacity.
Fetch fewer high-rate chunks for the same playback time.
Choose lower-rate versions of the next video chunks.
Keep high-rate chunks but shrink their displayed size.

Show answerHide answer

Save this lesson to see this answer.

Question 10 of 10

A standards proposal changes how an internet transport protocol communicates. What body develops internet protocols through open RFC work?

A regional internet registry
A domain-name registrar
The Internet Governance Forum
The Internet Engineering Task Force

Show answerHide answer

Save this lesson to see this answer.

Stuck on your own question? Homework Help

Key Terms

120 Key Terms, each with its definition

A

A and AAAA records

An A record stores an IPv4 address for a name, while an AAAA record stores an IPv6 address for a name.

A dual-stack service can publish both record types so clients can reach it using either address version.

Acknowledgment

A cumulative TCP acknowledgment reports the next byte expected after all earlier stream bytes have arrived without a gap.

Receiving bytes 1–700 in order permits ACK 701, even if those bytes were spread across several segments.

Adaptive streaming

Adaptive streaming selects among differently encoded media chunks using measured throughput, changing quality to fit delivery conditions.

A video player can request lower-rate chunks during congestion to reduce the chance of playback stopping.

Address scarcity

Address scarcity is limited availability of assignable address space, especially IPv4, encouraging reuse, transfers, and migration to IPv6.

A provider can use NAT to stretch existing IPv4 resources while gradually adding IPv6 connectivity.

Authentication

Authentication checks a claimed identity or endpoint; HTTPS verifies the server for the requested domain through certificates and key proof.

A valid connection to a store’s domain verifies the endpoint without guaranteeing that the store behaves honestly.

Authoritative server

An authoritative server holds a domain’s DNS records and provides answers for that domain from the information it serves.

A TLD referral directs a resolver to an authoritative server that can supply the domain’s address record.

Autonomous system

An autonomous system (AS) is a network or group of networks under one organization’s routing policy, identified by an AS number.

A university network and an ISP network can exchange routes as independently operated autonomous systems.

B

Bandwidth

Bandwidth is a link’s data-carrying capacity, measured in bits per second; it differs from the rate a transfer actually achieves.

A 100 Mbit/s connection can have lower useful throughput when several downloads share its capacity.

Best effort

Best effort means a network attempts delivery without guaranteeing that packets arrive, remain ordered, or appear only once.

An IP packet can be lost while TCP later recovers its bytes, preserving the application’s ordered stream.

BGP

Border Gateway Protocol (BGP) exchanges routing information between autonomous systems, with policies and business relationships shaping choices.

A provider can prefer a commercially agreed route even when another possible route is geographically shorter.

Bottleneck

A bottleneck is the capacity constraint that limits a path’s delivered rate, often its slowest or most heavily shared link.

Upgrading a server link leaves download speed unchanged if the user’s access link remains the limiting part.

Browser cache

A browser cache stores downloaded resources on a user’s device so usable copies can be reused without a complete new download.

A revisited page can display a stored logo while requesting other resources that need fresh copies.

Buffer

A buffer is temporary memory holding packets or other data that must wait.

A busy router stores arriving packets in its output buffer until that link is ready to send them.

C

Certificate

A certificate is a signed public document linking a name to a public key, checked during server authentication in secure connections.

The browser checks whether the certificate covers the requested host name before accepting the server.

Certificate authority

A certificate authority (CA) signs certificates that link names to public keys under a trust system used by clients such as browsers.

A trusted signature helps validate a certificate, but the server must still prove it holds the matching private key.

Checksum

A checksum is a computed value used to detect many kinds of damage to received data.

A TCP segment with a failed checksum is discarded so its data can be recovered through retransmission.

Circuit switching

Circuit switching reserves a fixed share of capacity along a path for a connection, keeping that share assigned during pauses.

A silent moment in a traditional reserved telephone circuit does not release its share to another call.

Client and server

A client requests a service, while a server provides it; the same host can take different roles in different exchanges.

A laptop is a client when fetching a report and a server when supplying a shared file to another host.

CNAME and NS records

A CNAME record makes one name an alias for another; an NS record identifies an authoritative name server for a domain.

An alias can lead a resolver to another host name, while an NS referral points it toward a source of DNS answers.

Concentration

Concentration places substantial infrastructure or service provision with a few organizations, bringing scale and shared dependencies.

Separate sites may fail together when they rely on the same large cloud provider during its outage.

Congestion control

TCP congestion control adjusts sending to network conditions so competing connections do not continually overwhelm busy links.

A sender slows down when loss indicates congestion even if the receiving host still has free buffer space.

Content delivery network

A content delivery network (CDN) serves distributed copies of resources to reduce distant transfers and share request load.

A nearby CDN server can provide a video chunk without every viewer contacting one distant origin server.

A cookie is data a server asks a browser to store and return on applicable later requests, often to identify a session.

Returning a session cookie lets a shop associate a later cart request with the same logged-in session.

D

Datagram

A datagram is an individually handled message unit, such as a UDP message, that can travel without an ordered-stream guarantee.

Two UDP datagrams can arrive in a different order from the one in which the sender transmitted them.

Decapsulation

Decapsulation removes protocol wrappers and passes their contents up the receiving stack.

A receiving host removes the link wrapper before its IP layer examines the enclosed network packet.

Default gateway

A default gateway is the local router a host normally uses to reach destinations outside its own network.

A laptop sends an overseas server’s packets first to its home router, without changing the remote IP destination.

DHCP

Dynamic Host Configuration Protocol (DHCP) automatically leases network settings such as an IP address, gateway, and DNS resolver.

A newly connected laptop can obtain usable local settings without its user typing an address manually.

Digital divide

The digital divide is unequal access to useful digital connectivity, shaped by infrastructure, affordability, skills, and other barriers.

A rural network’s coverage helps little for households that cannot afford the equipment or service price.

DNS time to live

A DNS time to live (TTL) limits normal cache reuse of a record, measured in seconds.

A recently changed address can remain old in a resolver’s cache until that entry’s remaining lifetime ends.

DNSSEC

DNSSEC adds signatures and validation to DNS so a resolver can check signed data’s authenticity without encrypting the query itself.

A validating resolver can reject an invalid signed answer even though observers may still read a classic DNS query.

Domain Name System

The Domain Name System (DNS) is a distributed directory that maps hierarchical names to records such as addresses and mail servers.

A browser obtains an address for www.example.org before connecting to a server at that address.

Domain registry

A domain registry operates a top-level domain and its registration system, distinct from companies that sell registrations to the public.

The registry for a TLD supports the registered names under it while registrars handle individual customer sales.

E

Email authentication

Email authentication uses checks such as SPF, DKIM, and DMARC to help assess whether a claimed sending domain is authorized or authentic.

A message can pass domain-related checks while still being unwanted, so passing them is not a guarantee of useful mail.

Encapsulation

Encapsulation wraps higher-layer data in a lower layer’s control information.

TCP bytes become an IP payload, and the resulting IP packet becomes the payload of a local link frame.

Encryption

Encryption transforms readable data into a protected form that requires suitable keys to recover, limiting access to the content.

HTTPS encryption can hide a page’s contents while an observer still sees packet sizes and transmission timing.

End-to-end encryption

End-to-end encryption protects message contents with keys held at the intended endpoints, keeping relaying servers from reading them.

A provider can store and forward an encrypted chat message without holding the keys needed to read its text.

F

Flow control

TCP flow control limits outstanding data according to receiver-advertised buffer capacity, protecting the receiving endpoint from overload.

A slow application can make its host advertise less receive space even while the network path is fast.

Forwarding

Forwarding selects a packet’s next outgoing link and sends it toward the next device.

A router forwards a packet by looking up its destination prefix in the table already available to it.

Forwarding table

A forwarding table maps destination prefixes to outgoing links or next-hop information.

A specific prefix entry can send a packet on one link while a default entry sends other destinations elsewhere.

Fragment

A URL fragment is the part introduced by a hash sign that a client handles locally, often to identify a position within a document.

In https://www.example.com/guide#ports, the browser handles #ports without sending it in the HTTP request.

Fragmentation

Fragmentation is the separation of the global internet into increasingly disconnected or differently controlled network spaces.

Different national restrictions can reduce cross-border reachability and raise concerns about a splinternet.

Frame

A frame is a link-layer unit carrying data and control information across one link.

A router replaces an incoming Ethernet frame with a new wrapper appropriate for its outgoing link.

H

Header

A header carries control information, such as addresses, for a protocol’s data unit.

A router examines the IP header to choose a next hop without interpreting a photograph carried in the packet’s payload.

Host

A host is an endpoint device that runs applications and exchanges data over a network.

A smart sensor is a host when its application sends readings across a network to a monitoring service.

HTML

Hypertext Markup Language (HTML) describes web documents and their structure, including links and references to other resources.

The browser reads HTML before it requests an image referred to by the page and draws the resulting document.

HTTP

Hypertext Transfer Protocol (HTTP) defines application requests and responses for resources, with methods, headers, statuses, and optional bodies.

A browser sends a GET request, then interprets the status and content returned by the responding server.

HTTP body

An HTTP body carries optional content, such as submitted form data or a returned page.

A successful document response can contain HTML in its body, while its headers describe that content.

HTTP header

An HTTP header carries request or response metadata, such as content type or cookies.

The response header Content-Type: image/png tells the client that the body is a PNG image.

HTTP method

An HTTP method states the action requested on a resource, such as GET for retrieval or POST for submitting data to be processed.

A form submission can use POST to carry entered values in the request body for the server to handle.

HTTP status code

An HTTP status code is a three-digit response value describing an outcome, with the first digit identifying its general class.

A 503 response reports service unavailability even though the server was reachable enough to send a response.

HTTP/3

HTTP/3 carries HTTP exchanges over QUIC, combining web request semantics with UDP-based reliable and secure transport.

A site using HTTP/3 can deliver a complete document without using a TCP connection for that exchange.

HTTPS

HTTPS is HTTP protected by TLS, providing encryption, integrity checks, and server authentication for a connection to a domain.

A submitted form can be hidden from network observers while its destination IP address remains visible.

I

IANA functions

The IANA functions maintain the DNS root zone, central address and AS-number pools, and shared protocol-parameter registries.

A globally consistent protocol-number registry lets independent implementations interpret the same numeric codes.

ICANN

The Internet Corporation for Assigned Names and Numbers (ICANN) coordinates top-level DNS policy and related identifier responsibilities.

A proposal for a new generic top-level domain belongs to DNS coordination, not to a local router’s forwarding work.

IETF

The Internet Engineering Task Force (IETF) develops open internet protocol standards through community participation and RFC publication.

Engineers can implement a common transport standard after participating in or reading the IETF’s technical work.

IMAP

Internet Message Access Protocol (IMAP) provides access to server-held mailboxes and synchronizes messages and state across devices.

Reading a message on a phone can update its read state when the same mailbox is opened on a laptop.

Integrity

Integrity protection detects unauthorized alteration of data in transit, helping the receiver reject changed protected messages.

TLS detects tampering with a protected response instead of letting the browser silently accept changed contents.

Internet

The internet is a global network of independently operated networks that exchange packets using shared protocols, especially IP.

A phone and a distant university server can communicate even though several operators carry their traffic.

Internet exchange point

An internet exchange point (IXP) is shared infrastructure where independent networks connect to exchange traffic directly.

Two providers can exchange local traffic at an IXP instead of paying a third provider to carry it onward.

Internet Governance Forum

The Internet Governance Forum (IGF) is a UN forum for multistakeholder internet-policy discussion that does not make binding decisions.

Participants can debate access and privacy at the IGF without the meeting itself enacting national law.

Internet service provider

An internet service provider (ISP) connects customers’ access networks to other networks and carries their traffic onward.

A household’s fiber line reaches an ISP, which exchanges traffic with other providers and content networks.

IP address

An IP address identifies a network interface for packet delivery; it can change and can be shared indirectly through translation.

A laptop receives a different network address after it leaves a home network and joins a workplace network.

IP time to live

IP time to live (TTL) limits router hops; routers reduce it and drop packets at zero.

Traceroute probes with increasing TTL values to discover responding routers at successive points on a path.

IPv4

IPv4 uses 32-bit addresses written as four decimal groups, each ranging from 0 to 255.

The address 192.0.2.15 has four groups, each representing one byte of its 32-bit address.

IPv6

IPv6 uses 128-bit addresses written in hexadecimal, providing a larger address space than IPv4.

An IPv6 address such as 2001:db8::25 uses a double colon to stand for omitted all-zero groups.

J

Jitter

Jitter is variation in packet delay that produces uneven arrival timing.

A live audio player may receive several packets close together after a pause and need to smooth their timing.

L

Latency

Latency is the elapsed time for data to travel from one endpoint to another.

A distant server can have high latency even when the connection can carry a large file at a high rate.

Link

A link carries bits between devices through copper, optical fiber, or radio.

A phone’s Wi-Fi radio link reaches the access point; later links can carry its traffic through fiber.

Longest prefix match

Longest prefix match selects the matching route with the most specified network bits.

A matching /25 entry takes priority over a matching /24 entry, even if the /24 entry appears first in the table.

M

MAC address

A MAC address is a link-layer identifier used for delivery on a local network; common Ethernet and Wi-Fi forms contain 48 bits.

A router uses local link addressing for the next device while the packet’s IP address names the remote endpoint.

Multistakeholder governance

Multistakeholder governance involves governments, businesses, civil society, technical communities, and academics in policy development.

A policy discussion can include users’ advocates alongside network operators and government representatives.

MX record

An MX record names a mail server that accepts messages for a domain and includes a preference used in choosing among its mail servers.

The sending server looks up example.net’s MX records before delivering a message to that domain.

N

Net neutrality

Net neutrality is the principle that ISPs should treat lawful traffic equally without favoring particular sites, apps, or content.

A proposal to slow a competing video service raises a neutrality question about equal network treatment.

Network address translation

Network address translation (NAT) rewrites addresses and often ports, allowing private IPv4 devices to share a public address.

A router’s translation table directs two incoming replies to the correct private devices behind one public address.

O

Overhead

Overhead is control or management information carried in addition to application data.

Sending more small packets increases the number of headers that must be carried for the same file.

P

Packet

A packet is a unit of network data containing a control header and a data payload.

A photo crosses a link as many packets, allowing packets from other conversations to take turns too.

Packet loss

Packet loss occurs when a packet fails to arrive, including when a full buffer causes a drop.

A download can still finish correctly after packet loss if its transport protocol recovers the missing data.

Packet switching

Packet switching shares links among arriving packets on demand, allowing several conversations to take turns using capacity.

A link carries one user’s page data during another user’s pause, improving use of intermittent demand.

Payload

A payload is the data carried inside a protocol unit, apart from its own control information.

If a 1,500-byte packet has 40 header bytes, only the remaining 1,460 bytes are available for its payload.

Peer-to-peer

Peer-to-peer (P2P) communication lets participating hosts take both client and server roles, requesting and supplying data to peers.

A file-sharing peer can upload pieces it has already obtained while downloading different pieces from others.

POP3

Post Office Protocol version 3 (POP3) retrieves mailbox messages, typically downloading them to one device for local reading.

A mail client can download messages through POP3, unlike SMTP’s role in sending them between servers.

Port

A port is a numbered transport endpoint used to direct data to a service or conversation.

A browser can connect from a temporary client port to HTTPS port 443 while maintaining other connections.

Prefix length

A prefix length counts the leading network bits, written after a slash in address notation.

The /24 in 203.0.113.0/24 leaves eight of an IPv4 address’s 32 bits outside the network prefix.

Private address

A private address is reserved for private networks and is not publicly routed.

Two homes can both assign 192.168.1.20 internally without claiming the same globally routed endpoint.

Propagation delay

Propagation delay is signal travel time along a link, calculated from distance divided by the signal’s speed in that medium.

A fiber journey across an ocean has a travel-time minimum that a higher bit rate cannot remove.

Protocol

A protocol defines message formats, exchange order, and participants’ actions.

HTTP gives a browser and server common rules for making a resource request and interpreting its response.

Protocol layer

A protocol layer performs a defined networking job and provides a service to the layer above while relying on the one below.

TCP can carry a browser’s bytes without the browser choosing how each Wi-Fi frame is transmitted.

Q

Query

A URL query is the part introduced by a question mark that carries parameters for the resource request sent to the server.

In /search?q=networks, the server receives a query parameter specifying the requested search term.

Queuing delay

Queuing delay is time a packet waits for an outgoing link while other packets use it, varying with the amount of competing traffic.

A call can become delayed during a large upload because packets wait behind traffic at a busy link.

QUIC

QUIC is a secure transport running over UDP that adds reliable streams, congestion control, and integrated encryption for applications.

HTTP/3 uses QUIC to deliver web resources reliably even though its underlying transport packets use UDP.

R

Recursive resolver

A recursive resolver obtains DNS answers on a device’s behalf, reusing cached information or following referrals through DNS servers.

An ISP resolver can answer a repeated lookup from its cache without asking a root server again.

Regional internet registry

A regional internet registry (RIR) distributes internet address blocks and AS numbers to providers and organizations in its region.

A provider seeking address resources deals with its regional registry rather than a company selling domain names.

Registrant

A registrant holds a domain registration for a renewable period under its terms.

A small publisher is the registrant of its chosen domain even when a separate company hosts its website.

Registrar

A registrar sells and manages domain registrations for people and organizations.

An organization renews its domain through its registrar so that its registration continues for another term.

Relay

A relay is an intermediate server that passes application traffic between endpoints, including when direct communication is unavailable.

A call may use a relay when NAT prevents its two devices from establishing a direct media path.

Retransmission

Retransmission sends data again after loss or detected corruption, using timers or acknowledgment patterns to trigger recovery.

TCP can resend the missing middle portion of a stream while keeping the later bytes already received.

RFC

A Request for Comments (RFC) is a numbered publication in the internet technical document series, including many protocol standards.

RFC 9293 documents TCP, allowing independent implementers to build compatible transport software.

Round-trip time

Round-trip time (RTT) measures the time for an outward and return network exchange.

If a request must wait for a reply before continuing, each additional exchange can add another RTT.

Router

A router forwards IP packets between networks using their destination addresses.

A home router sends an off-network request toward the provider rather than keeping it on the home network.

Routing

Routing determines paths and builds the information used to forward packets.

After a link fails, routing updates can make routers use another available path for later packets.

S

Segment

A TCP segment is a transport unit containing a TCP header and part of a byte stream; its boundary need not match an application message.

One long HTTP message can be carried in several TCP segments without changing its meaning to the browser.

Sequence number

A TCP sequence number identifies a position in the transmitted byte stream, allowing received pieces to be ordered and gaps found.

Bytes arriving after a missing range can be buffered until the earlier positions are recovered.

SMTP

Simple Mail Transfer Protocol (SMTP) submits outgoing email to a server and transfers messages between sending and receiving mail servers.

An outgoing mail server uses SMTP to deliver a queued message after DNS identifies the destination mail server.

Stateless protocol

A stateless protocol handles exchanges without automatically remembering earlier requests; applications can add their own stored state.

HTTP requests can carry a cookie that links them to server-held login state despite HTTP’s stateless design.

Statistical multiplexing

Statistical multiplexing shares capacity according to changing demand, allowing intermittent traffic sources to use the same link.

Many users who send only briefly can share a link, though overlapping bursts can still create queues.

Store-and-forward

Store-and-forward is the process of receiving an entire packet at a switch or router before sending it onto the next link.

A router must receive the final bits of a packet before beginning that packet’s next transmission.

Switch

A switch forwards link-layer frames among devices within a local network.

An office switch connects wired computers so that a frame can reach another device on the local network.

T

TCP

Transmission Control Protocol (TCP) provides a reliable ordered byte stream between endpoints using connections, acknowledgments, and recovery.

A file download can recover missing bytes without asking the application to rebuild every lost packet.

Three-way handshake

A TCP three-way handshake opens a connection through SYN, SYN-ACK, and ACK messages that establish the endpoints’ starting state.

A fresh TCP-based web connection needs setup before ordinary application data begins flowing.

Throughput

Throughput is the rate at which data is actually delivered to a receiver, constrained by bottlenecks, sharing, and overhead.

A server with a fast link cannot deliver through a slower home access line above that line’s capacity.

TLS

Transport Layer Security (TLS) protects a connection’s contents from reading and alteration and supports authentication of its endpoints.

A fresh secure web connection negotiates TLS before the browser sends its protected HTTP request.

Top-level domain

A top-level domain (TLD) is a named level directly below the DNS root, such as com or jp.

In mail.example.org, org is the TLD and mail is a subdomain within the registered domain.

Transit

Transit is a service in which one network pays another provider to carry its traffic onward to other internet networks.

A small ISP can buy transit from a larger provider to reach destinations beyond its direct connections.

Transmission delay

Transmission delay is the time needed to place a packet’s bits onto a link, calculated as packet bits divided by link rate.

The same packet takes longer to transmit over a lower-rate link even when the physical cable length is unchanged.

U

UDP

User Datagram Protocol (UDP) carries separate transport messages with ports and a checksum, without built-in delivery recovery or ordering.

An application receiving live updates can decide that a late old message should be skipped rather than recovered.

URL

A Uniform Resource Locator (URL) identifies a resource and how to reach it, including a scheme, host, and path with optional parts.

The URL https://www.example.com/help gives the secure scheme, server host name, and requested path.

W

W3C

The World Wide Web Consortium (W3C) develops web standards such as CSS and accessibility guidelines; WHATWG maintains HTML.

A web-accessibility requirement can draw on W3C guidelines without involving internet address allocation.

Web API

A web application programming interface (API) lets software request or submit information through web exchanges, often returning data.

A phone’s weather app can request JSON forecast data without displaying a conventional HTML web page.

World Wide Web

The World Wide Web is an internet application made of linked resources identified by URLs and retrieved through HTTP exchanges.

A browser uses the web to fetch a linked article, while a separate mail application also uses the internet.

Z

Zero-rating

Zero-rating excludes selected services from a data allowance while rivals’ traffic may count.

A subscriber can watch a favored app without using the data allowance while another app consumes the remaining data.

About this lesson

By StudeliPublished November 14, 2026Report a mistake

Keep learning

More Computer Science courses

Before thisBinary and Data RepresentationConvert between decimal, binary and hexadecimal, and explain how computers represent integers, real numbers, text, images and sound as bits.In Introduction to Computer Science Course

Parallel Processors and MulticoreCompare instruction-, data- and thread-level parallelism, and explain how multicore processors, GPUs and clusters share work and keep memory coherent.In Computer Organization and Architecture Course

Condition Variables and SemaphoresCoordinate threads with condition variables and semaphores, solve producer–consumer and readers–writers problems and prevent deadlock and other concurrency bugs.In Operating Systems Course

Routing and the Control PlaneCompute routes with link-state and distance-vector algorithms and explain how OSPF, BGP and SDN controllers build forwarding tables.In Computer Networks Course

Make your own Study Materials

Turn your notes, slides and documents into Study Guides, Practice Questions, Flashcards and 6 more tools.

Get Started