The TCP/IP 4-Layer Model

The OSI model is the textbook framework. The TCP/IP model is what the internet actually uses. They overlap a lot — TCP/IP just merges some OSI layers because in practice they’re rarely separated.

The 4 layers

TCP/IP Layer	OSI Equivalent	Examples
Application	L5, L6, L7	HTTP, DNS, SSH, SMTP, TLS
Transport	L4	TCP, UDP, QUIC
Internet	L3	IP, ICMP, ARP
Link	L1, L2	Ethernet, Wi-Fi, PPP

Why fewer layers

The OSI Session and Presentation layers were never widely implemented as separate things. Apps just handle their own session state and encoding. So TCP/IP folds them into Application.

Similarly, Physical and Data Link are usually so tightly coupled (a Wi-Fi card does both) that splitting them adds no value.

Encapsulation in TCP/IP

HTTP request                            ← Application layer
[TCP header | HTTP request]             ← Transport (TCP)
[IP header | TCP segment]               ← Internet (IP)
[Ethernet header | IP packet | trailer] ← Link (Ethernet)
electrical signals                      ← physical media

Each layer treats everything above it as opaque data. IP doesn’t care if you’re carrying TCP or UDP. Ethernet doesn’t care if you’re carrying IPv4 or IPv6.

The most important property: layered abstraction

A web app developer writing HTTP doesn’t think about routing. A network engineer configuring BGP doesn’t think about HTTP. This separation is what lets the internet scale — each team works at their layer.

Common confusion: “where does HTTPS fit?”

TLS sits between TCP and HTTP. In strict OSI it’s L6. In TCP/IP it’s part of the application layer. You’ll hear both. Don’t fight it — engineers use whichever fits the conversation.

Where to use which model

• Use OSI for teaching, troubleshooting (it’s the universal language), certifications.
• Use TCP/IP for actual implementation, RFCs, real-world reasoning.

What to learn next

Understanding bandwidth, latency, and throughput — three things people confuse — is up next.