Book summary “Computer Networks”: summary

Introduction

You live in a world where autonomous computing devices demand instant communication, yet physical channels remain inherently unreliable and prone to noise. When you send data across a distance, you face the core problem of transforming fragile analog signals into organized, secure digital messages that arrive intact. Without a systematic design, every application would need to reinvent routing, error correction, and security from scratch.

Andrew S. Tanenbaum addresses this complexity through the foundational principle of layered architectures. You learn to break down monumental communication tasks into manageable, independent protocol layers. Each layer performs a specific subset of functions, hiding lower-level details from the layers above it. This structured approach transforms chaotic electronic signals into dependable global services.

The book guides you upward through the entire network stack, starting from the hardware foundation and culminating in user-facing applications. You begin by examining physical media and modulation schemes, then ascend through data link protocols, medium access control, and network routing. You continue by exploring transport reliability and application layer services, finishing with the cryptographic mechanisms required to secure modern communications.

As you navigate this architectural landscape, you will find answers to several fundamental networking questions.

• How do physical transmission media carry raw bits across long distances?
• What mechanisms ensure reliable data delivery when underlying links drop frames?
• How do routers determine the optimal path for packets across massive interconnected systems?
• Why do transport protocols manage congestion and connection states in different ways?
• How do application protocols and security layers protect your data from interception?

1. Introduction to Network Architecture

A computer network connects a collection of autonomous computing devices that exchange information over various transmission media. You encounter this distributed architecture across modern workplaces and homes, replacing the obsolete model of a single centralized computer center.

Computer networks connect autonomous devices to share resources across layered protocol stacks.

Devices are interconnected if they exchange data using media like copper wire, fiber optic cable, or radio waves. These systems organize functions into a vertical sequence of levels to reduce design complexity. Each level offers services to upper layers while hiding implementation details through clean interfaces.

Network architecture relies on specific structural divisions:

• peers communicate virtually across machines using established rules;
• an interface separates the service specification from its internal implementation;
• a protocol stack coordinates operations across five layers in the illustrated network architecture.

Two philosophers communicate via translators who speak Dutch and secretaries who use faxes. This analogy illustrates how distinct layers pass messages down through interfaces before transmission occurs across physical channels.

Reference models standardize these interactions for global interoperability. The Transmission Control Protocol/Internet Protocol reference model was developed for the ARPANET with a core focus on seamless internetworking and network survivability during hardware failures.

ARPANET connected hundreds of universities and government installations using leased telephone lines. This robust design omits session and presentation layers while organizing functions into four distinct functional tiers.

The TCP/IP stack contains specific operational components:

• the link layer handles host-to-network connectivity;
• the internet layer uses the Internet Protocol and Internet Control Message Protocol;
• the transport layer features two end-to-end transport protocols, namely the Transmission Control Protocol and User Datagram Protocol;
• the application layer supports user services.

2. The Physical Layer Fundamentals

The physical layer transmits raw bits over guided media and wireless channels using modulation and multiplexing.

Transmission Media

Guided transmission media use physical paths to direct signals between communicating nodes. Copper wires provide traditional connectivity, but fiber optics offer vastly superior performance for high-speed transmission. Fiber optics use total internal reflection within ultra-thin glass fibers to transmit high-bandwidth light pulses over long distances with minimal attenuation.

When light rays hit the boundary between the glass core and cladding at or above a critical angle, they are entirely reflected internally instead of escaping. Single-mode fibers restrict light propagation to a straight line without bouncing, allowing much longer transmission distances than multimode fibers.

Key physical properties govern optical fiber performance:

• 10 microns - Core diameter threshold for single-mode fibers.
• 50 microns - Typical core diameter for multimode fibers.
• 100 Gbps - Data rate transmissible by single-mode fibers over 100 km without amplification.

The Theoretical Basis for Data Communication

Physical channels impose strict limitations on how fast you can push data across a network. Harry Nyquist proved in 1924 that a noiseless channel with a finite bandwidth has a strict maximum data rate limit. A signal filtered through a low-pass filter of bandwidth B can be completely reconstructed with exactly 2B samples per second. For a signal with V discrete levels, the maximum data rate is 2B log2 V bits per second.

This relationship produces clear operational limits for simple transmission lines:

• 6000 bps - maximum rate for a noiseless 3-kHz channel transmitting binary signals.

Random thermal noise fundamentally limits the maximum data capacity of any noisy channel. Claude Shannon published his foundational paper in information theory in 1948, extending capacity formulas to noisy environments by incorporating the Signal-to-Noise Ratio, or SNR. The maximum data capacity depends on the bandwidth and the Signal-to-Noise Ratio measured in decibels, or dB, representing a theoretical upper bound that real systems cannot exceed without changing bandwidth or noise characteristics.

Original ADSL operating over normal telephone lines with roughly a 1 MHz bandwidth and a 40 dB SNR achieved around 13 Mbps.

Real systems approach these mathematical bounds by choosing appropriate modulation techniques and managing channel impairments.

3. The Public Switched Telephone and Cellular Networks

Telecommunication infrastructure evolved from analog circuit switching to digital cellular packet networks and broadband cable.

The Public Switched Telephone Network

Network switching relies on two fundamentally different paradigms: circuit switching, which reserves a dedicated physical path and charges by time, and packet switching, which uses store-and-forward routing and charges by traffic volume. Circuit switching requires an end-to-end setup phase and guarantees fixed bandwidth without congestion during the call. It wastes resources if idle and fails entirely if a switch crashes. Packet switching dynamically routes independent packets, offering fault tolerance and high efficiency at the cost of queueing delays and per-packet congestion. Long-distance calls can experience 10 seconds of potential elapsed time for circuit setup alongside a 5 milliseconds per 1000 km signal propagation time delay.

Almon B. Strowger, a 19th-century Missouri undertaker whose competitor's wife was a telephone operator, invented automatic circuit-switching equipment to prevent calls intended for his funeral home from being misrouted. Strowger gear eliminated human operators from call routing.

Cellular Networks

Cellular systems achieve high capacity by dividing geographic regions into cells and reusing the same transmission frequencies in non-adjacent cells. Each cell uses a specific set of frequencies and is controlled by a base station at its center. When an area becomes overloaded, cells can be split into smaller microcells to allow more frequent reuse of frequencies and lower transmitter power. Advanced systems coordinate through a Mobile Switching Center, or MSC. AMPS systems featured typical cell sizes across 10 to 20 km.

Telephone companies create temporary microcells using portable towers with satellite links at sporting events and rock concerts.

GSM uses a combination of frequency division duplexing and time division multiplexing over 200-kHz channels to support multiple simultaneous connections. Each 200-kHz frequency pair is split by TDM into time slots structured into complex framing hierarchies. A mobile device does not transmit and receive simultaneously, allowing idle slots to be used for measuring signal quality from neighboring base stations through Mobile Assisted HandOff, or MAHO. Authentication uses a subscriber identity module, or SIM card.

GSM operations involve exact specifications:

• 200 kHz width of a GSM channel compared to 30 kHz for AMPS;
• 270,833 bps gross rate of each GSM channel divided among eight users;
• 13 kbps speech payload after error correction;
• 4.615 msec transmission period allowed for a data frame sharing the channel with seven other stations.

4. The Data Link Layer Design Issues

The data link layer encapsulates network layer packets into frames and provides services ranging from unacknowledged connectionless to acknowledged connection-oriented delivery.

You rely on this layer to transform a raw transmission facility into a line free of undetected transmission errors. The layer adds a header, payload field, and trailer to packets. It offers services depending on the medium characteristics:

• unacknowledged connectionless service for low error rates or real-time traffic;
• acknowledged connectionless service for unreliable links like WiFi;
• connection-oriented service with frame numbering and ordering for long and unreliable circuits.

Ethernet provides unacknowledged connectionless service, whereas 802.11 WiFi uses acknowledged connectionless service.

Error Detection and Correction

Error-detecting codes such as parity, Internet checksums, and CRCs are more efficient than error-correcting codes on low-error media when combined with retransmissions. Long wired cables often exhibit a bit error rate of 10^-6 per bit error rate for long wired cables, while typical LAN bit error rates reach 10^-10 typical LAN bit error rate.

Interleaving rearranges data transmission order across rows and columns to convert isolated error detectors into burst error detectors. Internet checksums use one's complement sums, while Cyclic Redundancy Checks, or CRCs, use polynomial modulo 2 division to catch burst errors and odd numbers of flipped bits.

IEEE 802 uses a 32 bits length for the IEEE 802 standard CRC polynomial standard CRC polynomial originally derived from Ethernet. When you transmit a block size of 1000 bits block size in error rate comparisons, these algebraic checks catch alterations before data reaches the network layer.

Elementary Data Link Protocols

A 1-bit sequence number is mathematically sufficient for a stop-and-wait ARQ protocol because ambiguity only exists between a frame and its immediate predecessor or successor.

Because the arrival of an acknowledgement for frame m implies that the prior frame has been safely received and acknowledged, the sender only ever chooses between the current frame and the next one. Alternating sequence numbers between 0 and 1 using modulo 2 arithmetic allows the receiver to detect duplicates and correctly pace unidirectional transmission. You track state variables such as MAX_SEQ, next_frame_to_send, and frame_expected to manage the exact 1 bit sequence number range (0 to 1) required for protocol 3.

5. The Medium Access Control Sublayer

Broadcast channels connect all users to everyone else, but transmitting simultaneously leads to collisions. The medium access control sublayer manages access to shared broadcast communication channels to allow multiple senders to share them fairly and efficiently.

Medium access control protocols regulate channel allocation in shared broadcast networks to prevent collisions.

A conference call where multiple participants speak simultaneously leads to chaos, similar to unmanaged multiaccess channels. You find these random access channels operating in local area networks and wireless systems.

Multiple Access Protocols

Slotted ALOHA divides time into discrete slots corresponding to frame transmission times, doubling channel capacity compared to pure ALOHA. By requiring stations to wait for the beginning of the next slot to transmit, the vulnerable period is halved. Throughput is defined as $S = Ge^{-G}$, which peaks at a load of $G = 1$ with a maximum channel utilization of 1/e, or approximately 36.8 percent.

Roberts proposed dividing time into discrete intervals in 1972, a method that was later revived for allocating shared channels when cable internet access was invented.

CSMA with Collision Detection allows stations to abort transmissions immediately upon detecting a collision, saving time and bandwidth. Station hardware listens while transmitting; if the read signal differs from the output signal, a collision is recognized.

In the worst-case propagation scenario across two farthest stations taking time $\tau$, a station must transmit for $2\tau$ without hearing a collision to be certain it has seized the channel. You encounter specific propagation properties in these topologies:

• 5 microseconds approximate propagation time $\tau$ on a 1-kilometer-long coaxial cable
• $2\tau$ duration a station must transmit without collisions to ensure it has seized the channel
• 1000000 times weaker received signals compared to transmitted signals in wireless setups

Ethernet

Switched Ethernet replaces shared hubs with internal switches, isolating collision domains and significantly increasing total system throughput. While hubs electrically connect all ports into a single collision domain requiring CSMA/CD, switches forward frames only to the specific destination port using a high-speed backplane. This isolation allows simultaneous frame transmissions and eliminates collisions on full-duplex links.

Hardware configurations feature standard installation limits:

• 4 to 48 ports in typical switch port counts
• 100 meters maximum cable run from hub or switch using twisted pairs
• 200 meters maximum cable run using high-quality Category 5 twisted pairs

6. Wireless LANs and Data Link Switching

Wireless networks manage hidden terminals and mobility while data link switches interconnect LANs using transparent bridging.

Wireless LANs

Because wireless devices operate half-duplex, you cannot transmit and listen for noise bursts simultaneously to detect collisions directly. Received signal strength can be a million times weaker than transmitted signal strength. The 802.11 MAC sublayer uses CSMA/CA combined with physical and virtual channel sensing to prevent collisions.

The Distributed Coordination Function uses exponential backoff and a Network Allocation Vector to track when the channel is logically busy.

Data Link Layer Switching

Bridges utilize a backward learning algorithm combined with a hash table and flooding to dynamically discover destination ports without manual configuration. When first plugged in, bridges use flooding for unknown destinations. By inspecting source addresses in incoming frames, they learn which machines are accessible on specific ports and periodically purge old entries. Station A sends a frame to station D; bridge B1 records where source machines are reachable and forwards frames accordingly.

Redundant physical links between bridges can create infinite looping cycles, which are eliminated by running a distributed spanning tree algorithm. Bridges exchange configuration messages to select a root based on unique global identifiers, construct shortest paths, and disable ports that create cycles to form a loop-free topology. Radia Perlman invented the spanning tree algorithm after being given a week to solve the loop problem.

Virtual LANs allow network administrators to partition physical LANs into multiple logical topologies in software to improve security, manage load, and control broadcast traffic. Standardized as IEEE 802.1Q, VLANs use VLAN-aware switches that inject tags into Ethernet frames to specify color identifiers, letting administrators group users by organization rather than physical location. Separating Human Resources department records from public web servers within the same physical office infrastructure requires these software controls. Hardware adjustments accommodate the protocol additions:

• 1518 bytes - previous maximum frame length limit
• 1522 bytes - increased maximum frame length limit to accommodate VLAN tags
• 0x8100 - VLAN protocol ID value

7. The Network Layer and Routing Algorithms

The network layer transfers packets across multiple interconnected networks. You must choose between connectionless datagram services and connection-oriented virtual circuits to manage this traffic.

The network layer routes packets across interconnected networks using distance vector and link state algorithms.

Network Layer Design Issues

In a datagram network, packets are injected individually without advance setup. Each packet contains the full destination address for independent routing. Each router maintains an internal table mapping possible destinations to outgoing lines. As network conditions change, routers can dynamically update their routing tables and send subsequent packets via different paths. A message is broken into four packets at the network layer. Packets 1, 2, and 3 follow path A-C-E-F, but packet 4 is sent via router B because A updated its routing table due to a traffic jam.

Virtual-circuit networks establish a fixed route from the source to the destination router prior to data transmission. Packets carry a short connection identifier instead of full addresses. Routers store internal tables mapping incoming connection identifiers to outgoing lines and new identifiers. Routers must be able to replace connection identifiers in outgoing packets to resolve numbering conflicts between different hosts. Host H1 establishes connection 1, and later H3 attempts to establish connection 1, forcing router A to assign a different connection identifier to the outgoing traffic to avoid a conflict at router C.

Routing Algorithms in a Single Network

Dijkstra's algorithm computes the shortest path between a source and all destinations using tentative and permanent labels. The algorithm builds a graph where nodes are routers and edges are communication lines with non-negative weights. It iteratively evaluates adjacent nodes, updates path lengths, and makes the smallest tentative label permanent until all shortest paths are found. The process utilizes specific parameters for network calculations:

• 1024 maximum number of nodes
• 1000000000 infinity larger than every maximum path

Link state routing distributes network topology information to all routers so they can independently compute shortest paths. Each router discovers neighbors, sets cost metrics, builds a link state packet, and floods it to all routers. Reliable flooding depends on sequence numbers and ages. IS-IS and OSPF are widely used link state protocols inside large networks and the Internet. A 32-bit sequence number takes 137 years to wrap around at one packet per second. This mechanism ensures that old packets do not corrupt routing tables during operation.

8. Congestion Control and Internetworking

Traffic management prevents network congestion while IP internetworking connects heterogeneous systems with global addressing.

Traffic Management at the Network Layer

Leaky and token bucket algorithms regulate bursty data flows, smoothing transmission rates to match network capacity and prevent buffer overruns. A leaky bucket enforces a constant outflow rate regardless of bursty input. A token bucket allows short-term bursts up to a capacity limit by requiring accumulated tokens to send packets. Service providers police customer traffic against agreed service level agreements using these mechanisms. A host generating a burst of 16,000 KB at 1000 Mbps with an average rate of 200 Mbps can be regulated using a token bucket with rate R equal to 200 Mbps and capacity B equal to 9600 KB.

Software-Defined Networking separates the control plane from the data plane to enable flexible, centralized network management. Traditionally, routing logic was vertically integrated into networking hardware. SDN separates the software and logic that select routes from the hardware responsible for packet lookups and forwarding. Figure 5-43 illustrates the architectural separation of the control plane and the data plane in SDN environments.

Internetworking

Internetwork routing uses a two-level hierarchy separating internal intradomain protocols from external interdomain protocols across autonomous systems. Autonomous systems managed by independent operators use interior gateway protocols internally and a single exterior gateway protocol like BGP across the broader internet to preserve administrative privacy and scale efficiently. IPv4 addresses use hierarchical prefixes to reduce routing table sizes, allowing large address blocks to be divided into internal subnets. A 32-bit address uses subnet masks so routers can forward packets based solely on the network prefix.

• A university with a /16 prefix splits its block into a /17 for Computer Science;
• another division creates a /18 for Electrical Engineering;
• a further split provides a /19 for Art.

Classless InterDomain Routing reduces routing table expansion in core routers by combining multiple smaller prefixes into larger aggregate blocks. Core routers keep about 300,000 prefixes without aggregation, but route aggregation reduces router tables to approximately 200,000 prefixes. Cambridge, Oxford, and Edinburgh address allocations are aggregated by a London router into a single 194.24.0.0/19 prefix before advertising to New York. This aggregate covers 8192 IP addresses starting at 194.24.0.0.

9. The Transport Layer and Reliable Communication

The transport layer provides end-to-end communication services, connection management, and reliable byte streams.

You rely on transport protocols to bridge the gap between network layer datagram routing and application requirements. Berkeley sockets provide a flexible and widely used set of primitives for Internet programming, abstracting transport services into reliable byte streams or datagrams. Sockets were first released as part of Berkeley UNIX 4.2BSD software distribution in 1983. The socket API allows applications to create endpoints, bind addresses, listen for connections, and use standard read and write system calls. New protocols like SCTP and QUIC introduce modifications to better handle groups of related streams.

Elements of Transport Protocols

Establishing connections requires careful coordination to handle delayed duplicates safely. The three-way handshake solves the problem of verifying connection requests and rejecting delayed duplicate segments safely. The initiator sends a connection request with sequence number x. The receiver replies with an acknowledgment of x and its own initial sequence number y. The initiator finally acknowledges y in its first data segment.

The Internet Transport Protocols: TCP

TCP establishes connections using a three-way handshake and releases them through independent simplex closures mapped onto an 11-state finite state machine. The client sends a SYN segment, the server responds with a SYN+ACK, and the client finishes with an ACK. Timers during the TIME WAIT state run for twice the multiplier of the maximum packet lifetime. To mitigate SYN flood attacks that exhaust server resources, hosts can employ cryptographic SYN cookies. In the 1990s, malicious senders crippled many Web servers by executing SYN flood attacks where they sent streams of SYN segments without completing the handshake.

TCP controls network congestion by implementing an Additive Increase Multiplicative Decrease control law using a congestion window and packet loss signals. TCP maintains an ack clock to smooth out traffic matching the slowest link path. When connections start or recover from losses, TCP uses slow start exponential growth until reaching a slow start threshold, after which it switches to linear additive increase. In the mid-1980s, the growing Internet experienced congestion collapse where goodput dropped precipitously by more than a factor of 100 before Van Jacobson implemented modern congestion control.

• TCP Tahoe
• TCP Reno
• TCP NewReno

10. The Application Layer Services

The application layer supports user-facing network services including the Domain Name System, email, web protocols, and content delivery.

The Domain Name System (DNS)

Name resolution uses a combination of stub resolvers, local recursive resolvers, and authoritative name servers, supported by extensive caching to improve performance. A stub resolver queries a local recursive resolver, which performs iterative queries up the naming hierarchy until it reaches an authoritative name server that holds the correct record. All responses and partial answers are cached with a Time_to_live field to prevent outdated entries. Each DNS query and response includes a 16-bit transaction identifier. A computer at cs.vu.nl querying for cs.uchicago.edu caches the result, allowing subsequent queries from the same domain to skip redundant steps.

The World Wide Web

HTTP uses a simple ASCII request-response model over TCP featuring explicit request methods and three-digit status codes to categorize transaction outcomes. Clients send requests starting with specific methods like GET, POST, or PUT, and servers return responses containing status codes divided into five groups ranging from informational 1xx to server error 5xx responses. HTTP/1.1 is specified in RFC 2616. RFC 2109 deals with cookies, while RFC 2965 specifies a more recent cookie standard largely rejected by industry.

SMTP is a simple ASCII protocol used by mail servers over TCP connections to reliably transfer messages between machines. Operating as a client-server protocol over port 25, SMTP uses four-character commands and numerical response codes to announce senders, verify recipients, and transmit message data. Because basic SMTP lacks native authentication and encryption, the ESMTP extension mechanism was introduced. A sample dialog shows a client sending a message from alice@cs.uchicago.edu to bob@ee.uwa.edu.au using commands like HELO, MAIL FROM, RCPT TO, and DATA.

Content Delivery Networks

Content Delivery Networks deploy distributed nodes globally using tree structures and DNS redirection to serve content from nearby locations with high performance. Pioneered by Akamai in 1998, CDNs place equipment racks with computers and disks inside ISP networks. They utilize DNS redirection to map client IP addresses to the nearest available replica, minimizing round-trip times and bypassing network congestion. Akamai operates 200,000 servers across more than 1500 networks and 120 countries. Akamai, Cloudflare, Limelight, and Dyn operate large CDNs, while providers like Netflix run their own CDN called Open Connect.

11. Network Security and Cryptography

Network security relies on three core properties known as the CIA triad, supplemented by authentication and nonrepudiation. Confidentiality prevents unauthorized users from reading data. Integrity ensures messages are not maliciously modified. Availability prevents service disruptions.

Network security safeguards communications through cryptographic algorithms, authentication protocols, firewalls, and secure transport layers.

Additional properties include authentication to verify identities and nonrepudiation to prevent senders from denying sent messages.

Cryptography

The RSA algorithm provides secure public-key encryption and digital signatures based on the mathematical difficulty of factoring large composite numbers into prime components. Invented by Rivest, Shamir, and Adleman, RSA computes public and private keys from the product of two large prime numbers, $p$ and $q$. Because factoring large numbers requires immense computational effort, recovering the private key remains infeasible. Rivest, Shamir, and Adleman received the 2002 ACM Turing Award for creating the RSA algorithm. RSA has survived break attempts for 40 years. The minimum key length recommended for RSA for good security is 2048 bits.

Management of Public Keys

A Public Key Infrastructure uses a hierarchical tree of authorities, ending in a set of trusted roots preloaded into browsers, to verify certificate chains without a single point of failure. Relying on a single global CA would create a severe bottleneck, a single point of failure, and high vulnerability to private key leakage. Instead, PKIs use a hierarchical structure where roots certify regional authorities or second-level CAs, allowing users to verify a chain of trust back to a trusted anchor. Modern browsers have the public keys of 100 roots preloaded. Alice verifies Bob's public key by checking the certificate signed by CA 5, then validates CA 5 using a certificate from RA 2, and finally trusts the root whose public key is built into her browser.

Transport Layer Security

Transport Layer Security establishes secure, authenticated, and encrypted connections between sockets above the transport layer. TLS uses a connection establishment subprotocol to negotiate parameters, authenticate the server via certificate chains, exchange a premaster key, and derive session keys for subsequent encrypted and hashed data transmission. During SSL connection establishment, the client chooses a 384-bit premaster key. In SSL data transmission, the maximum size of message units before compression and encryption is 16 KB.

Conclusion

Computer networks operate as complex hierarchical systems where every layer builds upon the services and abstractions provided by the layers beneath it.

You navigate network design and troubleshooting most effectively by tracing the strict boundaries between physical transmission, packet routing, reliable transport, and application services. The OSI reference model and the TCP/IP reference model demonstrate that modular decomposition remains the primary method for managing protocol complexity.

1. Separate network functions into distinct layers so that changes in physical medium or hardware do not require alterations to application software.
2. Implement robust error detection and flow control at the data link layer to prevent corrupted frames from propagating into upper routing processes.
3. Manage shared medium access carefully in broadcast environments to minimize collisions and maintain predictable throughput under heavy loads.
4. Utilize dynamic routing algorithms that adapt to topology changes and network failures without creating routing loops or broadcast storms.
5. Deploy congestion control mechanisms at the transport layer to prevent network collapse when aggregate traffic demands exceed router capacity.
6. Enforce end-to-end security through cryptographic primitives, proper key management, and robust authentication protocols at every architectural level.
7. Design applications to leverage existing naming services, reliable transport streams, and standard caching mechanisms rather than building custom protocols.

When you analyze a connectivity failure, isolate the problem by moving systematically up or down the protocol stack. Check the physical cabling or wireless signal strength first, verify framing and local medium access second, inspect routing tables and IP configurations third, test transport layer port availability fourth, and examine application headers last. By respecting these architectural boundaries and applying the core principles of multiplexing, framing, congestion avoidance, and security, you build resilient networks capable of scaling from local office environments to global internetworks.

10 Key Ideas

1. Network architecture organizes communication functions into a vertical protocol stack that hides lower-level details through clean interfaces.
2. The maximum data capacity of any noisy channel is fundamentally limited by its bandwidth and Signal-to-Noise Ratio as defined by Claude Shannon.
3. Medium access control protocols regulate channel allocation in shared broadcast networks to prevent data collisions.
4. Data link switches interconnect local area networks dynamically using transparent bridging and backward learning algorithms.
5. The network layer routes packets across interconnected networks using distance vector and link state algorithms.
6. The Transmission Control Protocol establishes reliable connections using a three-way handshake and controls network congestion through an additive increase multiplicative decrease control law.
7. The Domain Service Name resolves names using stub resolvers, local recursive resolvers, and authoritative name servers supported by extensive caching.
8. The RSA algorithm provides secure public-key encryption based on the mathematical difficulty of factoring large composite numbers into prime components.
9. A Public Key Infrastructure uses a hierarchical tree of authorities ending in a set of trusted roots preloaded into browsers to verify certificate chains.
10. Transport Layer Security establishes secure, authenticated, and encrypted connections by negotiating parameters and deriving session keys for data transmission.