What Happens When the Internet Goes Down? Ad-hoc Networks, Mesh, and Antifragile Communication

In the first hours after the February 6, 2023 earthquake, most cellular networks in the affected region collapsed. Base stations were damaged, electricity was cut, backbone lines severed. Millions of people were trapped — both under rubble and in the darkness of lost communication. This is not an exception; it is a pattern that repeats with every major disaster, cyberattack, or deliberate infrastructure disruption.

This article addresses one question: When the internet and cellular networks completely collapse, how do you communicate with the devices you have?

Why Cellular Networks Collapse

A cellular network doesn't collapse from a power cut alone — several links in the chain must break.

Base station: Each base station runs on local power and typically has 4-8 hours of UPS (uninterruptible power supply) backup. If power remains cut for more than 8 hours, base stations begin shutting down.

Backbone connection: Base stations connect to the core network via fiber or microwave links. When this connection breaks, even a physically intact base station cannot operate.

Core network: If the data centers housing core network components like HSS, MME, and SGW are damaged, all service stops — even if base stations are physically operational, no one can reach anyone.

Traffic surge: At the moment of a disaster, the entire affected population reaches for their phones simultaneously. The number of concurrent calls a base station can handle is finite; a sudden demand spike pushes the network far beyond capacity. This effectively collapses the network even when infrastructure is intact.

What Is an Ad-hoc Network?

A network formed by devices connecting directly to each other without needing centralized infrastructure (base station, router, server) is called an ad-hoc network. "Ad-hoc" is Latin for "for this purpose" — a structure built for an immediate need, not permanent.

The simplest form of an ad-hoc network is two devices communicating directly over Wi-Fi or Bluetooth. Its more complex form is MANET (Mobile Ad-hoc Network): a topology where moving devices dynamically discover each other, each device acts as both an endpoint and a router, and the network self-organizes without central coordination.

Mesh Network: Extending Coverage Without Infrastructure

A mesh network is the scaled form of an ad-hoc network. Each node carries coverage one step further. Even if two endpoints cannot see each other directly, the chain of nodes between them relays the message.

Hop count and latency: Each hop adds latency. A typical BLE mesh hop produces 10-50 ms of delay; in a 5-hop chain, latency can reach 50-250 ms. Acceptable for text messaging; borderline for voice calls.

Routing algorithms:

• Flooding: Each node forwards every received message to all neighbors. Simple but inefficient — redundant transmissions multiply as the network grows.
• AODV (Ad-hoc On-demand Distance Vector): Performs route discovery only when needed; doesn't burden nodes with no active connections.
• Gossip Protocol: Messages are forwarded to randomly selected neighbors; probabilistic but efficient in large networks.
• Delay-Tolerant Networking (DTN): In networks with intermittent connectivity (two nodes occasionally meeting in the same area), messages are stored and carried. It is the digital equivalent of the "message courier" model.

Technology Comparison: What Works When?

Bluetooth Mesh (Bridgefy, etc.)

Strong when: High device density. In festivals, concerts, and urban centers, mesh forms quickly within 100-meter range.
Insufficient when: Low density — if there are insufficient nodes between two devices, mesh cannot function. Large file transfer is not feasible.
Security: Bridgefy now uses Signal Protocol, though metadata is observable at each hop.

Wi-Fi Direct / IBSS (Ad-hoc Wi-Fi)

Strong when: High bandwidth over short distance. Can reach 100-300 Mbps between two devices. Suitable for file sharing and video transfer.
Insufficient when: Connection remains P2P between two devices; no automatic mesh routing.
Range: 50-200 meters in open areas.

LoRa / Meshtastic

Strong when: Low device density, wide area. Covers kilometers in rural, mountain, and remote terrain. Battery lasts days.
Insufficient when: Low data rate (max ~50 kbps). Voice, video, and large files cannot be carried. Ideal only for short text and location data.
Security: AES-128 encryption; verified in open-source Meshtastic.

Briar (Tor + Bluetooth + Wi-Fi)

Strong when: Maximum privacy priority. Works over Bluetooth/Wi-Fi without internet; switches to Tor when internet returns. Zero centralized infrastructure.
Insufficient when: Tor's slowness (5-30 s initial connection). Complex setup; not suitable for mass disaster use.

DTN: Carrying Data When There Is No Connection

Delay-Tolerant Networking is a network architecture designed for environments where connectivity is intermittent. NASA used this architecture for interplanetary communication; the same principles apply on Earth.

DTN's core principle is "store-carry-forward": when direct connection to a recipient is unavailable, a message is stored at an intermediate node. When that node moves into the same environment as the recipient, it delivers the message.

Practical scenario: A disaster has severed the connection between two neighborhoods. A motorcycle rescue team passes through both; their device collects messages accumulated at each neighborhood and delivers them to the other. The digital version is the software implementation of this physical "message courier" model.

Antifragility: Growing Stronger from Collapse

Nassim Taleb's antifragility concept states: merely resilient things are protected from stress; antifragile things grow stronger because of it.

Centralized architectures are fragile: when one node fails, everything dependent on it stops. Mesh networks are antifragile: when one node drops, others find new routes; total capacity may decrease but the network doesn't collapse entirely. As more nodes drop, remaining nodes make new routing decisions to operate more efficiently.

This difference is decisive in disaster communication design. Dozens of small, interconnected nodes instead of a single powerful antenna; distributed, peer-to-peer storage instead of a single data center.

Practical Protocol: What to Do in a Disaster

When communication infrastructure collapses, the right actions depend on available devices and the environment:

Prioritize text, not bandwidth: Voice calls and video congest disaster networks. Short text messages use fewer resources and are more reliably delivered.

Power management: Reduce screen brightness, close background apps, disable radios beyond Bluetooth/Wi-Fi. Every mAh matters.

Discover locally first: Connect with devices in the same area first. If one person reaches the internet, they can serve as a bridge for others.

Rendezvous point: Establish a predetermined location and frequency (for amateur radio users) or mesh application channel in advance.

Conclusion

Internet outages are not a hypothesis — they are a documented, recurring reality. Ad-hoc networks, mesh topologies, and DTN systems are the engineering responses developed against this reality. None of these responses is perfect in isolation: Bluetooth mesh depends on density, LoRa on range, Briar on privacy, DTN on timing. But when the right tool is selected for the right scenario, a handful of devices and the right software are sufficient to fill the void left by collapsed infrastructure. Antifragility is not building a powerful center — it is designing a periphery that makes the center's collapse irrelevant.