Wi-Fi covers a room; cellular covers a city. The gap is not just range — it is who owns the infrastructure, how the network is organized, and how a connection survives when you move. This topic completes the wireless story from net-wireless-fundamentals and ties back to the connection-migration ideas in net-quic-http3.

The Cellular Architecture

A cellular network splits into two broad regions:

• Radio Access Network (RAN) — the base stations that reach your phone over the air. In 5G these are gNodeB (gNB) sites; in 4G/LTE they were eNodeB (eNB).
• Core Network — the central switching and service functions that connect the RAN to the public internet (4G’s EPC; 5G’s cloud-native 5G Core).

Your phone connects to one base station’s cell. As you move, the network hands you between cells — this is handover (handoff).

Cells and Handover

• Cells are geographic coverage regions, each served by a base station with a portion of spectrum. Macro cells cover kilometers; small cells (pico/femto) fill indoor and dense urban dead zones.
• Handover transfers an active session from one cell to the next. Two flavors:
  • Hard handover — break before make (brief gap). Used where the radio can only talk to one tower at a time.
  • Soft handover — make before break (talk to both briefly), used in technologies like CDMA/3G to eliminate the gap.
• Handover decisions depend on signal strength, load, and interference; 5G and 4G keep the data path so TCP/QUIC sessions survive the jump.

4G vs 5G

5G’s headline is low latency (for autonomous driving, remote surgery, AR) and network slicing — carving one physical network into virtual, dedicated networks with independent QoS (e.g. a low-latency slice for vehicles, a high-bandwidth slice for video).

Wi-Fi Roaming vs Cellular Handover

• Wi-Fi roaming is client-driven and best-effort: you reassociate to whichever AP you reach in the same SSID (802.11r speeds this up), but neither the client nor AP coordinates a clean data handoff — a TCP flow can still stall.
• Cellular handover is network-controlled: the RAN coordinates with the core and neighboring cells to move the session before your signal degrades, so live calls and video streams survive.

The practical takeaway: mobile devices hop between Wi-Fi and cellular all day, and the network layer has to cope with the IP address change.

Connection Migration

A TCP connection is identified by its 4-tuple (src IP, src port, dst IP, dst port). The moment your device switches from Wi-Fi to cellular (or between cells with a new IP), the 4-tuple changes and a TCP connection dies — apps must reconnect. This is exactly the problem QUIC solves (recall net-quic-http3): QUIC carries a Connection ID independent of the IP, so a connection migration keeps the session alive across networks.

So the wireless data path is:

Application (HTTP/3) ── QUIC (Connection ID → survives migration)
  └─ UDP/IP (address may change)
      └─ Cellular/Wi-Fi link (handover keeps RF continuity)

Network Slicing (5G)

Slicing lets operators run multiple virtual end-to-end networks on one physical infrastructure:

• eMBB slice — high bandwidth for video/AR.
• URLLC slice — ultra-reliable low latency for industrial control.
• mMTC slice — massive machine-type communication for IoT.

Each slice has its own guaranteed latency, bandwidth, and isolation — a clean example of virtualization applied to a telecom network, mirroring the VMs and containers you meet in Operating Systems and DevOps.

Verify It Yourself

1. While on a video or VoIP call, switch your phone between Wi-Fi and cellular. Note whether the stream survives — if it does, that is QUIC connection migration; if it drops, you saw the TCP 4-tuple limit.
2. Observe signal bars change as you move — that is handover in action (the RAN moving you between cells).
3. Re-read the net-wireless-fundamentals “roaming” section and contrast client-driven Wi-Fi roaming with network-controlled cellular handover in a short mental model.