Learn
A2.1.5 | FUNCTION OF THE TCP/IP MODEL (HL ONLY)
01 | ONE REQUEST, SEVERAL DIFFERENT JOBS
Imagine opening a lesson on a school website. Your browser needs to ask for the correct resource, but that request also needs a route across networks and a way to cross the Wi-Fi link beside you. These are different problems. One enormous set of rules would be difficult to design, change and troubleshoot.
The TCP/IP model organises network communication into four layers: application, transport, internet and network interface. Each layer concentrates on a particular responsibility and uses services provided by the layer below.
A model is a way of describing how responsibilities fit together. It is not four physical boxes inside a computer, and it does not mean that every message must use TCP. UDP is also part of the TCP/IP protocol suite.
This is an HL-only objective. Your goal is to describe each layer and the interaction between layers, including where reliable delivery is supplied and where it is not.
02 | SEE THE FOUR LAYERS TOGETHER
Read down the stack for sending and up the stack for receiving. The application delegates delivery work instead of needing to control radio signals itself. At the destination, lower layers handle delivery details before the application interprets the message.
What does the service need to communicate?
Example: an HTTP request for a lesson
Which software endpoints communicate, and what delivery service is supplied?
Examples: TCP or UDP; transport ports
How is an IP packet addressed and forwarded across networks?
Example: IP; routers use routing information
How does data cross the current link?
Examples: Ethernet or Wi-Fi framing and physical transmission
In this four-layer version, the network interface layer includes link functions and physical transmission. Some other diagrams split these into two layers. For this IB objective, use the four names above consistently.
03 | APPLICATION: GIVE THE COMMUNICATION MEANING
The application layer contains protocols that define service-specific messages and their meaning. For a web service, HTTP defines requests such as GET and responses such as a status code with a resource. A browser understands these rules, while the layers underneath do not need to understand the lesson’s text.
Application layer does not simply mean “the program with a window”. DNS name resolution and DHCP configuration also use application-layer protocols, even when operating-system software performs them automatically.
For example, the message “request /lesson.html” belongs to the web service. Deciding whether that resource exists and constructing a response are application concerns. Choosing which Wi-Fi radio transmission carries the bytes is a different concern.
HTTPS is protected HTTP communication. In a common HTTP/1.1 or HTTP/2 arrangement, TLS works between HTTP and TCP. The four-layer model groups functions broadly; it does not require a separate named layer for every real protocol.
04 | TRANSPORT: COMMUNICATION BETWEEN SOFTWARE ENDPOINTS
A host can run a browser, a game and a file transfer at the same time. Reaching the correct host is not sufficient: received data must also reach the appropriate software endpoint. The transport layer uses logical port numbers as part of this identification. A transport port is not a cable socket.
Transport supplies a service between communicating endpoints. TCP supplies a reliable, ordered byte stream. UDP supplies separate datagrams without built-in reliable or ordered delivery. Both use IP underneath.
For TCP, application bytes can be carried across multiple segments, and received bytes are delivered as a stream. One application write is not guaranteed to become one segment or one receiver read. Applications that need message boundaries define those boundaries themselves.
A server might listen on TCP port 443 while a client uses a temporary local port. IP addresses identify the participating interfaces; the transport protocol and ports help identify the communication endpoints.
05 | INTERNET: ADDRESS AND FORWARD ACROSS NETWORKS
The internet layer carries transport data inside IP packets. IP headers contain source and destination IP addresses, enabling hosts and routers to handle delivery between networks.
A router examines destination addressing and routing information to select a next hop. A next hop is the next device on the path, not necessarily the final destination. The router does not normally need to interpret the HTTP request to make this forwarding decision.
IP provides a best-effort packet service. Packets can be lost, delayed or reordered; IP itself does not provide TCP’s reliable ordered byte-stream service. A route is also not necessarily fixed for an entire exchange.
The internet layer can be used inside a school network as well as across the public internet. The layer’s name describes its internetworking role, not a requirement to contact a public website.
06 | NETWORK INTERFACE: CROSS THE CURRENT LINK
An IP packet cannot travel as an abstract object. On an Ethernet link it is carried in an Ethernet frame; on a wireless link, link-specific framing and radio transmission carry it. These local delivery and transmission functions belong to the network interface layer in this model.
A frame is a link-layer unit containing link control information and a payload. A payload is the data being carried. An Ethernet frame can carry an IP packet as its payload. Ethernet source and destination MAC addresses concern delivery on the current link.
If a server is outside your subnet, your first Ethernet frame is usually addressed to your next-hop router’s local interface, not directly to the distant server’s MAC address. A basic Ethernet switch uses MAC information to forward frames locally.
Cables, radio signals and interface hardware matter here, but the layer also includes rules such as framing and access to the medium. It is more than simply “the cable”. Link error detection can identify damaged frames; recovery behaviour depends on the technology and does not replace end-to-end transport recovery.
07 | THREE IDENTIFIERS, THREE DIFFERENT QUESTIONS
| Identifier | What it helps identify | Worked example |
|---|---|---|
| IP address | Interface addressing for IP delivery across networks | Destination server: 203.0.113.20 |
| Transport port | Logical transport endpoint on a host | Destination TCP port: 443 |
| Link address, such as Ethernet MAC | Local link sender/recipient | First-hop destination: router’s LAN interface |
The example IP addresses in this lesson are reserved documentation addresses. They are not addresses to configure on your school network.
A useful question sequence is: which IP destination, which transport endpoint there, and which local next hop now? These identifiers cooperate. They are not interchangeable versions of the same address.
08 | ENCAPSULATION: CARRY ONE UNIT INSIDE ANOTHER
Encapsulation means wrapping data with the control information needed by another layer. A header is control information placed before a unit’s payload. Some link formats also add a trailer after the payload.
In a simplified TCP-over-IP-over-Ethernet example, TCP adds a TCP header to bytes being carried in a segment. IP carries the segment inside a packet. Ethernet carries the packet inside a frame. The physical link transmits a signal representing that frame.
An envelope analogy can help: the original note remains inside successive wrappers. However, real protocols can split, combine or transform data, so this is not a promise that one whole web page becomes one packet. Encryption and segmentation are omitted from the small example below.
Ethernet header: local link addresses
IP header: source and destination IP
TCP header: ports, sequence and other control fields
Part of a message being carried
Ethernet trailer: includes a frame check sequence
09 | BUILD THE WRAPPERS, THEN REMOVE THEM
Use the controls to follow an illustrative unencrypted HTTP exchange over TCP, IPv4 and Ethernet. It shows one small chunk of application data; handshake, encryption, segmentation and error cases are deliberately omitted.
At the receiver, decapsulation means processing control information and passing the contained data to the appropriate higher-layer protocol. It does not mean decrypting the data: removing delivery headers and decrypting protected content are different operations.
10 | A ROUTER CHANGES THE LINK WRAPPER
Suppose a client sends data through one router to a server on another Ethernet network. The first frame has the client’s link address as source and the router’s receiving interface as destination. Its enclosed IP packet is addressed to the server.
The router processes the incoming link frame, examines the IP destination and selects an outgoing interface and next hop. It then places the forwarded IP packet in a new link frame suitable for the outgoing link. The old Ethernet frame is not carried unchanged all the way to the server.
In this ordinary routing example without NAT, tunnelling or proxies, the source and destination IP addresses stay the same across the router. Some IP header fields do change, such as the IPv4 time-to-live value and associated header checksum. “The packet is completely unchanged” would therefore be inaccurate.
Try the two hops below. The MAC labels are symbolic, not literal hardware addresses. In a multi-router route, the outgoing link destination could be another router rather than the final server.
11 | HOW THE LAYERS SUPPORT RELIABLE COMMUNICATION
Different mechanisms solve different problems. A link can detect certain transmission errors; IP forwards packets across networks; TCP at the endpoints tracks byte positions, acknowledges progress and retransmits when required. The receiver can hold later bytes until a gap is filled.
For example, suppose TCP data for byte ranges 0–99 and 200–299 arrives but 100–199 is missing. TCP can wait for recovery of the gap before releasing the later bytes in order. IP does not reconstruct this ordered stream for the application.
TCP’s flow control helps avoid overwhelming the receiver. Congestion control responds to network conditions. Neither means the path can never fail: an unrecoverable problem can cause the connection to fail.
Reliability is not the same as encryption, authentication or successful completion of a business task. TCP does not encrypt a password, and a transport acknowledgement does not prove a document was saved. Application responses and appropriate security mechanisms address those separate concerns.
12 | THE MODEL ALSO WORKS WITH UDP
Replacing TCP with UDP changes the transport service, not the existence of all four layers. A DHCP message, for example, uses an application protocol over UDP, with IP and the local link supporting delivery.
UDP itself does not recover missing messages or reorder them into a reliable stream. An application or higher-level protocol can add its own mechanisms when needed. Therefore reliable transmission is a cooperative design choice, not a universal guarantee created by writing “TCP/IP” on a diagram.
For this objective, learn the four roles first. As an extension, HTTP/3 uses QUIC over UDP, with QUIC adding transport and security functions. This shows why a simple layer model is useful but cannot be treated as an exhaustive map of every implementation detail.
13 | OPENING A LESSON: PUT THE ROLES TOGETHER
Assume a laptop already has valid network configuration, knows the server address and uses an HTTP/1.1 connection protected by TLS over TCP. Those assumptions let us concentrate on the four layers rather than repeat every setup task.
The application constructs an HTTP request. TLS protects the communication; TCP carries the protected bytes with transport control information. IP supplies destination addressing for forwarding, and the local link carries the packet to the next hop. Routers repeat the forwarding and link-delivery work along the path.
At the server, link and internet processing deliver the data to the correct transport protocol. TCP supplies the ordered stream to the upper software, TLS processes protected records, and the web application interprets the HTTP request. The response travels through the same kinds of layers in the opposite direction.
A real browser may reuse an established connection, use cached data or choose HTTP/3. The worked case describes a specified arrangement, not a claim that every page view repeats an identical sequence.
14 | USE THE LAYERS TO INVESTIGATE A PROBLEM
Suppose a student says “the internet is broken”. The layer model gives you more useful questions. Is there a working link? Is IP configuration correct? Can traffic reach the destination? Is the intended service accepting transport connections? Did the application return an error?
A disconnected Ethernet cable points toward link and physical delivery. Incorrect routing information points toward IP delivery. A web server returning HTTP 404 has communicated an application response; it does not by itself show the Wi-Fi link failed.
Treat layers as a way to organise evidence, not as a machine that diagnoses the cause from one symptom. A timeout might involve a failed server, a routing problem, congestion, a firewall rule or several other causes. A successful test of one service does not prove every service works.
15 | A CLEAR FOUR-LAYER SUMMARY
| Layer | Describe the function | Examples / units |
|---|---|---|
| Application | Defines service messages and their meaning | HTTP, DNS, DHCP; application data |
| Transport | Provides communication between software endpoints; TCP supplies ordered reliable stream transport | TCP segments; UDP datagrams; ports |
| Internet | Provides IP addressing and packet forwarding between networks | IPv4/IPv6 packets; routers |
| Network interface | Handles delivery over the current link and physical transmission | Ethernet/Wi-Fi frames; link addresses and signals |
Avoid calling every unit a “packet” when the question asks you to distinguish layers. “Packet” can be used broadly in conversation, but TCP segment → IP packet → link frame makes encapsulation clearer. UDP uses datagrams instead of TCP segments.
Remember that layer names describe functions. A combined home router can perform switching, routing, wireless access and application services. One physical device need not correspond to exactly one layer.
16 | DESCRIBE FUNCTIONS AND CONNECT THE LAYERS
For describe, give the main characteristics and functions. A strong answer names each layer, states its responsibility and shows how it passes work to the next layer. Listing four names alone does not describe their function.
For example: “The application layer defines the service messages. The transport layer carries data between software endpoints, with TCP providing ordered reliable stream delivery. The internet layer adds IP addressing for packet forwarding across networks. The network interface layer carries the packet over the current link. The receiver processes these layers in reverse so that application data reaches the intended software.”
When reliability is asked about, add the specific TCP mechanisms rather than claiming that IP guarantees delivery. Use connectives such as because, therefore and so that to link a mechanism to its purpose.
Further reading: Internet host architecture, TCP, IPv4 and IP over Ethernet. These standards are extension references; the four-layer explanation above is the course focus.