If you picked up this book hoping to skip straight to building a mesh network on your roof, I respect that impulse. But I’m going to ask you to spend some time here first, because the fundamentals of networking are not merely academic prerequisites — they are the lens through which every design decision in an alternative network becomes intelligible.
Here is the thing about alternative networks: they break rules. They simplify protocol stacks, flatten hierarchies, invent new addressing schemes, and route packets in ways that would make a Cisco-certified engineer raise an eyebrow. But you cannot meaningfully break rules you don’t understand. Every departure from the standard model is a design decision with tradeoffs, and understanding those tradeoffs requires knowing what the standard model does and why it does it.
So we begin here, at the foundations. If you already have a networking background, this chapter will reframe what you know through the lens of decentralization and resilience. If you are new to networking, this chapter will give you a solid foundation without burying you in unnecessary detail. Either way, by the end, you will understand the building blocks from which every network — conventional or alternative — is constructed.
The Open Systems Interconnection (OSI) model is the conceptual framework that networking professionals use to think about how data moves from one device to another. Published by the International Organization for Standardization (ISO) in 1984, it divides the problem of networking into seven distinct layers, each responsible for a specific aspect of communication.
Think of it like sending a physical letter. You write the message (that’s the application layer). You put it in an envelope and address it (presentation and session). You hand it to the postal service, which decides how to route it (transport and network). The postal service loads it onto trucks and planes (data link and physical). At the other end, the process reverses: the letter is unloaded, sorted, delivered to the right address, and opened by the recipient.
The seven layers, from bottom to top:
The physical layer is the raw infrastructure — the actual medium through which bits travel. This includes copper cables, fiber optic strands, radio waves, and even light pulses through the air. The physical layer defines things like voltage levels, pin layouts, cable specifications, radio frequencies, modulation schemes, and data rates.
For alternative networks, the physical layer is where many of the most important choices are made. When you choose between Wi-Fi, LoRa, packet radio, or Ethernet, you are making a physical layer decision. Each medium has fundamentally different characteristics:
The physical layer choice constrains everything above it. You cannot stream video over LoRa. You cannot reach a village 5 km away with Ethernet (without repeaters). Understanding these constraints is the first step to designing a network that actually works.
The data link layer handles communication between devices that are directly connected — on the same wire, the same radio channel, the same network segment. It is responsible for framing (packaging bits into structured units called frames), error detection, and controlling access to the shared medium.
This layer is where MAC addresses (Media Access Control addresses) live — the 48-bit hardware addresses burned into every network interface. When your computer sends a frame on a local network, it addresses it to the destination’s MAC address. Switches operate at this layer, forwarding frames based on MAC addresses.
For alternative networks, the data link layer is where things get interesting. Standard Wi-Fi operates in infrastructure mode, where all traffic flows through a central access point — a fundamentally centralized design. Alternative networks typically use ad-hoc mode (also called IBSS) or mesh mode (802.11s), which allow devices to communicate directly with each other without any access point. Some mesh protocols, like BATMAN-adv (Better Approach To Mobile Ad-hoc Networking — Advanced), operate entirely at Layer 2, making the mesh transparent to higher layers. From the perspective of applications running on mesh nodes, it looks like all nodes are on the same local network, even if packets are hopping through multiple intermediaries.
The network layer handles routing — getting packets from a source to a destination that may be many hops away, across multiple networks. This is the layer where IP addresses live, where routers make forwarding decisions, and where the global internet’s routing infrastructure operates.
In the conventional internet, routing is hierarchical. Your device sends packets to your home router, which sends them to your ISP’s router, which sends them to a regional backbone, and so on up and down the hierarchy until they reach the destination. This hierarchy is efficient but creates the centralization and single points of failure that alternative networks seek to avoid.
Alternative network routing protocols — Babel, OLSR (Optimized Link State Routing), BMX6/7, CJDNS, Yggdrasil — operate at this layer but use fundamentally different approaches. Instead of hierarchical routing tables maintained by large organizations, they use distributed algorithms where every node participates in routing decisions. We will explore these protocols in depth in Chapter 7.
The transport layer provides end-to-end communication services for applications. The two most important protocols here are TCP (Transmission Control Protocol) and UDP (User Datagram Protocol).
TCP provides reliable, ordered delivery. It establishes a connection, ensures every packet arrives, retransmits lost packets, and delivers data in the correct order. This reliability comes at a cost: overhead from acknowledgments, retransmissions, and connection management.
UDP provides unreliable, unordered delivery. It sends packets and hopes they arrive, with no guarantees. This sounds inferior, but it has critical advantages: lower latency, lower overhead, and better suitability for applications where occasional packet loss is acceptable (voice, video, sensor telemetry, real-time gaming).
For alternative networks, the choice between TCP and UDP has profound implications. TCP was designed for the relatively stable, low-latency connections of wired networks. On mesh networks, where routes can change frequently, latency can be high and variable, and links can appear and disappear, TCP’s aggressive retransmission and congestion control algorithms can actually make things worse — a phenomenon sometimes called TCP meltdown. When a mesh route changes and packets are temporarily delayed, TCP interprets this as congestion, backs off its sending rate, and can take a long time to recover. UDP-based protocols, or TCP variants designed for high-latency and unstable links, often perform better in mesh environments.
The session layer manages sessions — ongoing connections between applications. It handles establishing, maintaining, and terminating communication sessions, as well as synchronization and recovery.
In practice, this layer is often considered somewhat vestigial in modern networking. Most of its functions are handled by application-level protocols or by libraries within applications themselves. For alternative networks, the session layer concept is still relevant: maintaining a session across a mesh network where routes are constantly changing requires careful design, and protocols like QUIC (which combines transport and session functionality) show how modern protocols are rethinking these traditional boundaries.
The presentation layer handles data formatting, encryption, and compression — translating between the format applications use internally and the format used for network transmission.
Like the session layer, this is often absorbed into application-level code in modern practice. However, its functions — especially encryption — are critically important for alternative networks. TLS (Transport Layer Security), which provides encryption for HTTPS and many other protocols, conceptually lives at this layer. In alternative networks, where traffic may traverse nodes operated by unknown parties, encryption at this layer (or above) is not optional — it is essential.
The application layer is where users interact with the network. Web browsers, email clients, chat applications, file transfer programs — all of these are application layer entities. Protocols like HTTP, SMTP, DNS, SSH, and MQTT operate at this layer.
For alternative networks, the application layer is where the most visible innovation happens. Decentralized applications like Matrix (federated messaging), IPFS (distributed file storage), Scuttlebutt (offline-first social networking), and Briar (mesh messaging) are all application layer protocols designed specifically for environments where the conventional internet’s assumptions — always-on connectivity, central servers, stable routes — do not hold.
The seven-layer model is a useful thinking tool, but alternative networks rarely implement it cleanly. In practice, they tend to collapse layers, skip layers, or add new functionality that does not fit neatly into any single layer:
The lesson is this: the OSI model is a map, not the territory. Use it to understand and communicate about networking concepts, but do not be surprised when real-world alternative networks deviate from it significantly. The question is never “does this conform to the model?” but “does this work for the use case?”
While the OSI model provides a useful conceptual framework, the internet does not actually run on it. The internet runs on the TCP/IP model (also called the Internet Protocol Suite), which was developed independently and predates the OSI model by several years.
The TCP/IP model has four layers:
This simpler model is more practical and maps more closely to how software is actually structured. When you write a networked application, you typically interact with the transport layer (opening a TCP connection or sending a UDP datagram) and let the operating system handle everything below.
TCP is the workhorse of the internet. Web browsing, email, file transfer, remote administration — the majority of internet traffic flows over TCP. Its key features:
For alternative networks, TCP’s congestion control is a double-edged sword. Algorithms like TCP Reno, CUBIC, and BBR are tuned for the characteristics of the commercial internet — relatively stable routes, relatively low latency, relatively predictable packet loss due to congestion. On a mesh network with rapidly changing routes, high and variable latency, and packet loss due to radio interference (not congestion), these algorithms can behave poorly.
Some alternative network deployments address this by:
UDP is TCP’s lean sibling. No connection setup, no acknowledgments, no retransmissions, no ordering guarantees. You send a datagram and move on. It either arrives or it does not.
This makes UDP ideal for:
For alternative networks, UDP is often the better choice at the application level, especially for the real-time and sensor applications that dominate many mesh and LoRa deployments. Many mesh routing protocols also use UDP (or raw IP) for their control traffic, since routing updates need to be fast and lightweight, and a lost update will simply be superseded by the next one.
In the conventional internet, IP addresses are allocated hierarchically by IANA (Internet Assigned Numbers Authority) to regional registries, then to ISPs, then to end users. This hierarchy enables efficient routing — a router in Tokyo does not need to know the specific location of every device in Brazil; it just needs to know which direction to send packets addressed to Brazilian IP blocks.
Alternative networks often cannot or do not want to participate in this hierarchy. They use IP addresses differently:
This cryptographic approach to addressing is a profound departure from conventional networking. In a traditional network, your IP address tells you where you are in the network topology. In CJDNS or Yggdrasil, your address tells you who you are — it is a cryptographic identity that does not change regardless of where you connect. This has enormous implications for mobility, security, and decentralization that we will explore in later chapters.
Network topology is the arrangement of nodes and links in a network — the shape of the network when you draw it on a whiteboard. Topology determines nearly everything about a network’s behavior: its resilience to failures, its performance characteristics, its scalability, and its cost. For alternative networks, topology is arguably the single most important design decision.
In a star topology, all nodes connect to a single central node (a hub, switch, or access point). All communication between nodes passes through this central point.
[Node A]
|
[Node B]--[HUB]--[Node C]
|
[Node D]
Advantages:
Disadvantages:
Real-world examples: Most home Wi-Fi networks are star topologies — your wireless router is the hub, and every device connects to it. Corporate networks with a central switch are star topologies. Most cell phone networks are star topologies at the local level, with each cell tower serving as the hub for all devices in its coverage area.
Relevance to alternative networks: Star topology is the opposite of what most alternative networks aim for. Its reliance on a central node makes it vulnerable to the exact failures and control points that alternative networks exist to avoid. However, small star networks often serve as building blocks within larger mesh or hybrid topologies — a household’s local Wi-Fi network (star) might connect to a neighborhood mesh network via its router.
In a ring topology, each node connects to exactly two other nodes, forming a circular chain. Data travels around the ring in one direction (or both, in a dual ring), passing through each node until it reaches its destination.
[Node A]---[Node B]
| |
[Node D]---[Node C]
Advantages:
Disadvantages:
Real-world examples: Token Ring networks (once common in corporate environments) used ring topology. Many metropolitan fiber networks (SONET/SDH rings) use dual-ring topology for resilience. Some industrial control networks use ring topology for its deterministic performance.
Relevance to alternative networks: Pure ring topology is uncommon in alternative networks, but the ring concept appears in some mesh routing protocols where certain nodes form backbone rings. The dual-ring concept — two independent paths between any pair of nodes — foreshadows the redundancy that mesh networks provide.
In a bus topology, all nodes connect to a single shared communication medium — a single cable, a single radio channel, or a single frequency. Every node can hear every transmission, and some mechanism (like CSMA/CD — Carrier Sense Multiple Access with Collision Detection) is needed to prevent simultaneous transmissions from colliding.
[Node A]---[Node B]---[Node C]---[Node D]
|
(shared bus)
Advantages:
Disadvantages:
Real-world examples: Original Ethernet (10BASE2, 10BASE5) used bus topology with coaxial cable. Many early local networks were buses. In wireless networking, a single Wi-Fi channel is essentially a shared bus — all devices on the same channel share the same medium and must take turns transmitting.
Relevance to alternative networks: While physical bus topology is largely obsolete, the concept is highly relevant because most wireless networks are logically a bus. All devices sharing a Wi-Fi channel or a LoRa frequency are sharing a bus, and the limitations of bus topology — especially contention and collision — apply directly. Understanding bus behavior helps you understand why mesh networks with many nodes on the same channel can experience performance degradation, and why channel planning and frequency management matter.
A tree topology is a hierarchy of star networks — stars connected to other stars, forming a branching structure like an upside-down tree.
[Root]
/ \
[Switch1] [Switch2]
/ \ / \
[A] [B] [C] [D]
Advantages:
Disadvantages:
Real-world examples: The global internet’s routing structure is broadly hierarchical — Tier 1 providers at the top, Tier 2 in the middle, end users at the bottom. Corporate networks are typically trees. The DNS system is a tree (root servers, TLD servers, authoritative servers).
Relevance to alternative networks: Tree topology represents exactly the hierarchical, centralized architecture that alternative networks seek to avoid. However, elements of tree structure appear in many alternative networks in practice — especially in community networks where a few high-capacity backbone nodes connect to many lower-capacity leaf nodes. The key difference is that alternative networks try to ensure that the failure of any single node (even a “root” node) does not isolate large portions of the network. This is where mesh topology enters the picture.
Mesh topology is the star of this book — the topology that makes resilient, decentralized, self-healing networks possible. In a mesh network, nodes are connected to multiple other nodes, creating redundant paths between any pair of devices.
There are two forms:
Full mesh: every node is directly connected to every other node. In a network of $n$ nodes, this requires $\frac{n(n-1)}{2}$ links.
[A]------[B]
|\ \ / /|
| \ \ / / |
| \ \/ / |
| \/\/ |
| /\/\ |
| / /\ \ |
| / / \ \ |
|/ / \ \|
[C]------[D]
Partial mesh: nodes are connected to some other nodes, providing multiple paths without the overhead of a full mesh.
[A]------[B]
| |
| [E] |
| / \ |
[C]------[D]
Advantages:
Disadvantages:
Real-world examples:
Why mesh is the most important topology for alternative networks:
Mesh topology embodies the philosophy of alternative networking. There is no central authority that controls the network. There is no single point that can be attacked, confiscated, or switched off. The network adapts to changes — nodes joining, leaving, moving, or failing — without human intervention. And perhaps most importantly, every participant strengthens the network. In a star topology, a new node adds load to the center. In a mesh topology, a new node adds capacity, paths, and resilience.
This does not mean mesh is always the right choice. For a small home network, a simple star is fine. For a backbone link between two buildings, a point-to-point connection is appropriate. But for any network that needs to be resilient, decentralized, and community-operated, mesh topology is the foundation.
Real-world networks almost never conform to a single pure topology. Instead, they combine elements of multiple topologies to balance performance, cost, resilience, and practicality. These combinations are called hybrid topologies.
A typical community alternative network might look like this:
Understanding the pure topologies is essential because it gives you the vocabulary and mental models to analyze and design these hybrid networks. When someone says “our community network has a single point of failure,” you can immediately think about what topological changes would eliminate it.
Every communication system needs a way to identify who is talking to whom. In networking, this takes several forms, operating at different layers and serving different purposes.
Every network interface — every Ethernet port, every Wi-Fi adapter, every Bluetooth chip — has a MAC address (Media Access Control address). This is a 48-bit address, typically written as six pairs of hexadecimal digits: 00:1A:2B:3C:4D:5E.
MAC addresses are assigned by manufacturers and are intended to be globally unique. The first three octets (24 bits) identify the manufacturer (OUI — Organizationally Unique Identifier), and the last three are assigned by the manufacturer to individual devices.
MAC addresses operate at Layer 2 and are used for communication on the local network segment. When your computer sends a frame to another device on the same Ethernet or Wi-Fi network, it addresses the frame to the destination’s MAC address.
For alternative networks, MAC addresses have several implications:
While MAC addresses identify hardware on a local segment, IP addresses identify devices across networks — they are the logical identity used for routing decisions.
IPv4 addresses are 32-bit numbers, written as four decimal octets: 192.168.1.100. This gives a theoretical maximum of about 4.3 billion addresses — a number that seemed enormous in the 1980s but was exhausted years ago. The internet continues to function through NAT (Network Address Translation), where many devices share a single public IP address, with the NAT device (typically your router) tracking which internal device each connection belongs to.
For alternative networks, IPv4’s relevant address ranges include:
IPv6 addresses are 128-bit numbers, written as eight groups of four hexadecimal digits: 2001:0db8:85a3:0000:0000:8a2e:0370:7334. This gives approximately $3.4 \times 10^{38}$ addresses — enough to assign a unique address to every atom on the surface of the Earth and still have plenty left over.
IPv6 is increasingly important for alternative networks for several reasons:
Several alternative network protocols are built exclusively on IPv6:
Subnetting is the practice of dividing a network into smaller sub-networks. A subnet mask (or prefix length in CIDR notation) specifies which portion of an IP address identifies the network and which identifies the host.
For example, 192.168.1.0/24 means the first 24 bits (192.168.1) identify the network, and the last 8 bits identify hosts within it — allowing 254 usable host addresses (256 minus network and broadcast addresses).
In alternative networks, subnetting serves several purposes:
A community mesh network might assign subnets like:
10.0.1.0/24 for the downtown mesh cluster10.0.2.0/24 for the university campus cluster10.0.3.0/24 for the residential neighborhood cluster10.0.0.0/24 for backbone infrastructureHumans think in names (example.com); computers think in numbers (93.184.216.34). The Domain Name System (DNS) bridges this gap, translating human-readable domain names into IP addresses.
The global DNS is a hierarchical, largely centralized system. At the top are 13 root server clusters (actually hundreds of servers distributed globally via anycast). Below them are TLD (Top-Level Domain) servers (.com, .org, .net, country codes). Below those are authoritative name servers for individual domains. When you type a URL into your browser, your device queries this hierarchy to resolve the name to an IP address.
This system works well but has significant problems for alternative networks:
Alternative networks use several approaches to naming and discovery:
mDNS allows devices to resolve names on the local network without any DNS server. Each device announces its own name, and other devices can query for names by sending multicast queries. If your device is named mycomputer, it is accessible as mycomputer.local on the local network. This is the technology behind Apple’s Bonjour and Linux’s Avahi.
mDNS works beautifully for small local networks and is natively supported by all major operating systems. It does not scale well beyond a single subnet, but for a household or small office alternative network, it provides naming with zero configuration.
DNS-SD builds on mDNS to allow devices to advertise and discover services. Instead of just names, devices announce what services they offer — web servers, file shares, chat servers, printers. This enables automatic discovery of services on an alternative network without any central directory.
For larger alternative networks, several distributed naming approaches exist:
For many practical alternative network deployments, a simple local DNS server (like dnsmasq or Pi-hole running on a Raspberry Pi) provides naming for the local network while forwarding external queries to the internet (when available). This hybrid approach gives the convenience of DNS without full dependence on the global DNS infrastructure.
In conventional networks, your IP address is tied to your location in the network topology. If you move your laptop from home to the office, you get a different IP address because you are on a different network.
Mesh networks challenge this assumption. Nodes may be mobile. The topology may change. Connections may come and go. Several approaches address this:
The trend in alternative networking is strongly toward cryptographic addressing and IPv6 autoconfiguration — approaches that require no central authority and work seamlessly as nodes join, leave, and move.
Routing is the process of determining the path a packet takes from source to destination. In a network with a single path — a simple star or bus — routing is trivial. But in any network with multiple possible paths (which includes every mesh network and most non-trivial networks), routing becomes a complex and critical function.
Static routing means manually configuring routing tables on each device. You tell each router: “to reach network X, send packets out interface Y.” This is simple, predictable, and requires no routing protocol overhead. It is also completely impractical for any network larger than a handful of nodes, and it cannot adapt to failures — if a link goes down, traffic to destinations reached via that link is simply lost until a human reconfigures the routes.
Dynamic routing means that routers automatically discover the network topology, calculate optimal paths, and adapt to changes. Dynamic routing protocols exchange information between routers, allowing them to build a consistent picture of the network and update their forwarding tables as conditions change.
For alternative networks, dynamic routing is not optional — it is the foundation. The entire point of a mesh network is that it adapts to changes without human intervention. When a node goes offline, routes are recalculated. When a new node joins, it is automatically integrated. When a link degrades, traffic shifts to better paths. All of this requires dynamic routing protocols.
Distance-vector routing protocols work on a simple principle: each router tells its neighbors how far it is from every destination it knows about. Neighbors use this information to determine the best path to each destination — choose the neighbor that reports the shortest distance, and send traffic that way.
The most famous distance-vector protocol is RIP (Routing Information Protocol), one of the oldest routing protocols on the internet. The algorithm is intuitive:
Distance-vector protocols are simple to implement and require little memory. However, they have a well-known problem: count-to-infinity. When a link fails, it can take a long time for the information to propagate through the network, with routers bouncing “I can reach X through you” messages back and forth, slowly incrementing the distance until it reaches an infinity value. Techniques like split horizon, poison reverse, and hold-down timers mitigate this but do not eliminate it entirely.
Babel, one of the most popular routing protocols for alternative networks, is a modern distance-vector protocol that incorporates techniques to avoid count-to-infinity problems and converge quickly in the dynamic topologies of mesh networks.
Link-state routing protocols take a different approach: each router broadcasts information about its directly connected links (their state — up or down, their cost) to every router in the network. Each router then has a complete map of the network topology and can independently calculate the optimal path to every destination using Dijkstra’s algorithm.
The most common link-state protocols are OSPF (Open Shortest Path First) and IS-IS (Intermediate System to Intermediate System), both widely used in enterprise and carrier networks.
Link-state protocols converge faster than distance-vector protocols and do not suffer from count-to-infinity. However, they require more memory (every router stores the entire network topology) and more bandwidth (link-state advertisements must be flooded to all routers).
OLSR (Optimized Link State Routing) is a link-state protocol designed specifically for mobile ad-hoc networks (MANETs). It uses multipoint relays (MPRs) — selected nodes that are responsible for forwarding link-state information — to reduce the flooding overhead that makes traditional link-state protocols impractical in large wireless networks.
Conventional routing protocols assume a relatively stable network. Links come up and stay up for weeks or months. Routes change infrequently. Convergence after a failure is measured in seconds, and failures are rare enough that this delay is acceptable.
Mesh and ad-hoc networks violate all of these assumptions:
These conditions have driven the development of specialized mesh routing protocols:
We will explore each of these protocols in detail in Chapter 7. For now, the key takeaway is that routing in alternative networks is fundamentally different from routing in conventional networks, and protocols designed for one environment often perform poorly in the other.
Three terms dominate any discussion of network performance: bandwidth, latency, and throughput. They are related but distinct, and understanding the differences is crucial for setting realistic expectations about alternative networks.
Bandwidth is the maximum rate at which data can be transmitted over a link — the theoretical capacity of the pipe. It is measured in bits per second (bps) and its multiples: kilobits (kbps), megabits (Mbps), gigabits (Gbps).
Some representative bandwidths:
| Technology | Typical Bandwidth |
|---|---|
| LoRa (long range) | 0.3–50 kbps |
| HF packet radio | 0.3–9.6 kbps |
| VHF/UHF packet radio | 1.2–9.6 kbps |
| Bluetooth Classic | 1–3 Mbps |
| Meshtastic (LoRa mesh) | 1–10 kbps |
| 802.11b Wi-Fi | 11 Mbps |
| 802.11n Wi-Fi | 150–600 Mbps |
| 802.11ac Wi-Fi | 433–6,933 Mbps |
| 802.11ax Wi-Fi (Wi-Fi 6) | Up to 9.6 Gbps |
| Gigabit Ethernet | 1 Gbps |
| Fiber optic | 10–100+ Gbps |
Note the enormous range. LoRa and fiber optic are separated by a factor of millions. This is not a deficiency of LoRa — it is a fundamental tradeoff. LoRa achieves its multi-kilometer range and micro-watt power consumption precisely because it uses very little bandwidth. Understanding these tradeoffs is essential for choosing the right technology for each part of an alternative network.
Latency is the time it takes for a packet to travel from source to destination. It is measured in milliseconds (ms) and has several components:
Some representative latencies:
| Path | Typical Latency |
|---|---|
| Same LAN (Ethernet) | < 1 ms |
| Same city (internet) | 5–20 ms |
| Cross-continent (internet) | 50–100 ms |
| Satellite (GEO) | 500–700 ms |
| Satellite (LEO, e.g., Starlink) | 20–50 ms |
| Wi-Fi mesh (3–5 hops) | 10–50 ms |
| LoRa mesh (3–5 hops) | 1–10 seconds |
| HF radio link | 100–500 ms |
| Delay-tolerant network | Minutes to days |
For alternative networks, latency is often higher than on commercial networks — sometimes much higher. A LoRa mesh message might take several seconds to reach its destination. An HF radio email might take hours. A sneakernet delivery takes days. This is not a bug; it is a deliberate tradeoff. Alternative networks optimize for resilience, coverage, and independence, accepting higher latency as the cost.
Throughput is the actual data rate achieved in practice — it is always less than bandwidth due to protocol overhead, contention, errors, retransmissions, and other real-world factors.
A Wi-Fi link rated at 300 Mbps might achieve 150 Mbps of actual throughput under ideal conditions, and considerably less in a busy environment. A mesh network where each hop shares a channel has even lower throughput: a rough rule of thumb is that throughput halves with each hop, since the same radio must both receive and retransmit. A 3-hop mesh path on a 54 Mbps link might achieve only 6–8 Mbps of end-to-end throughput.
For alternative networks, setting realistic throughput expectations is essential:
The conventional internet optimizes aggressively for speed. Fiber runs in straight lines between data centers. Content is cached at edge servers close to users. Routing protocols find the shortest path. Everything is designed to minimize the time between “I clicked a link” and “I see the page.”
Alternative networks make different choices. They optimize for:
These priorities necessarily come at the cost of raw performance. A community mesh network will never match the speed of a fiber-to-the-home connection. A LoRa mesh will never match the bandwidth of a cell phone network. The question is not “is this as fast as the commercial internet?” but “does this provide the connectivity I need, with the independence, resilience, and privacy I want?”
For many use cases — text messaging, email, file sharing, sensor monitoring, emergency coordination, local collaboration — the answer is an emphatic yes.
Most alternative networks are wireless — they use radio waves to communicate without physical cables. Understanding the basics of wireless communication is essential for designing, deploying, and troubleshooting these networks.
All wireless communication uses electromagnetic radiation — specifically, the radio frequency (RF) portion of the electromagnetic spectrum. Radio waves are characterized by three related properties:
Frequency ($f$): the number of oscillations per second, measured in Hertz (Hz). Radio frequencies range from about 3 kHz to 300 GHz. Higher frequencies carry more data but travel shorter distances and penetrate obstacles less effectively.
Wavelength ($\lambda$): the physical distance of one complete oscillation. Wavelength and frequency are inversely related: $\lambda = \frac{c}{f}$, where $c$ is the speed of light (approximately $3 \times 10^8$ m/s). A 900 MHz signal has a wavelength of about 33 cm; a 2.4 GHz signal has a wavelength of about 12.5 cm; a 5 GHz signal has a wavelength of about 6 cm.
Propagation: how radio waves travel through the environment. This depends on frequency, power, terrain, atmospheric conditions, and obstacles. Key propagation behaviors include:
Radio spectrum is a shared and limited resource, regulated by national authorities (the FCC in the United States, OFCOM in the UK, ETSI in Europe, and their equivalents worldwide). Most of the spectrum is licensed — assigned to specific users (broadcasters, mobile operators, military, emergency services) who have exclusive rights to use those frequencies in a given area.
However, some bands are designated for Industrial, Scientific, and Medical (ISM) use and can be used by anyone without a license, subject to power and usage restrictions. These ISM bands are the foundation of most alternative networking:
| Band | Frequency Range | Common Uses |
|---|---|---|
| 433 MHz | 433.05–434.79 MHz | LoRa (Europe/Asia), remote controls, sensors |
| 868 MHz | 868–870 MHz | LoRa (Europe), IoT devices, smart meters |
| 915 MHz | 902–928 MHz | LoRa (Americas), Meshtastic, ISM devices |
| 2.4 GHz | 2.400–2.4835 GHz | Wi-Fi, Bluetooth, Zigbee, microwave ovens |
| 5 GHz | 5.150–5.825 GHz | Wi-Fi (802.11a/n/ac/ax), some radar |
| 5.8 GHz | 5.725–5.875 GHz | Point-to-point links, FPV drones |
| 6 GHz | 5.925–7.125 GHz | Wi-Fi 6E, emerging uses |
| 60 GHz | 57–71 GHz | WiGig, short-range high-bandwidth links |
For alternative networks, the most important bands are:
An antenna is the device that converts electrical signals into radio waves (for transmission) and radio waves back into electrical signals (for reception). The antenna is one of the most important components in any wireless network — the right antenna can make the difference between a working link and no connection at all.
Key antenna parameters:
Common antenna types for alternative networks:
Omnidirectional antennas radiate in all directions (in the horizontal plane). They have relatively low gain (2–9 dBi) but cover 360°. Ideal for mesh nodes that need to communicate with neighbors in all directions. Most Wi-Fi routers use small omnidirectional antennas.
Sector antennas cover a wide angle (typically 60–120°) with moderate gain (10–17 dBi). Used at central locations that need to serve a wide area without full 360° coverage — like a community network access point mounted on a tower, covering a neighborhood.
Directional (dish or panel) antennas focus energy into a narrow beam (10–30° or less) with high gain (15–30+ dBi). Used for point-to-point links between distant locations. A pair of directional antennas aimed at each other can establish links of 10, 20, or even 100+ km with standard Wi-Fi equipment.
Yagi antennas provide moderate directional gain (10–18 dBi) in a relatively compact form factor. Named after Hidetsugu Yagi, who popularized the design in the 1920s, they are the familiar “TV antenna” shape. Useful for medium-distance links and for concentrating signal toward a specific area.
Homemade and DIY antennas are a proud tradition in alternative networking. Tin can “cantenna” waveguide antennas, biquad antennas made from copper wire, and even Pringles-can antennas have been used successfully in community networks. While commercial antennas generally perform better, DIY antennas can be remarkably effective and cost almost nothing.
For reliable wireless links, especially over long distances, understanding line of sight (LoS) and Fresnel zones is critical.
Line of sight means there is an unobstructed straight line between the transmitting and receiving antennas. For short-range links (within a building or a small yard), obstructions matter less because signals can reflect, diffract, and scatter around obstacles. But for long-range point-to-point links, a clear line of sight is usually essential.
However, having a clear line of sight is necessary but not sufficient. Radio waves do not travel in a perfectly thin line — they spread out in an elliptical zone called the Fresnel zone (pronounced “freh-NEL”). The first Fresnel zone is the most important; if more than about 40% of it is blocked by terrain, buildings, or vegetation, the signal will be significantly attenuated.
The radius of the first Fresnel zone at its widest point (the midpoint of the link) is:
\[r = 17.32 \sqrt{\frac{d}{4f}}\]where $r$ is the radius in meters, $d$ is the link distance in kilometers, and $f$ is the frequency in GHz.
For a 5 km link at 2.4 GHz, the first Fresnel zone radius at the midpoint is about 12.5 meters. This means the straight-line path between the antennas needs to be clear of obstacles for at least 12.5 meters above and below the line, and to each side. In practice, this is why long-distance wireless links require elevated antennas — mounted on rooftops, towers, or hilltops.
Signal attenuation — the loss of signal strength — occurs for many reasons:
For alternative network planners, a link budget calculation is essential before deploying any long-range link. A link budget adds up all the gains (transmitter power, antenna gains at both ends) and subtracts all the losses (cable losses, free-space path loss, obstacle attenuation, fade margin) to determine whether a link will work:
\[\text{Received Power} = \text{Transmit Power} + \text{TX Antenna Gain} + \text{RX Antenna Gain} - \text{Path Loss} - \text{Cable Losses} - \text{Misc. Losses}\]If the received power exceeds the receiver’s sensitivity (the minimum signal it can decode), the link will work. If not, you need more power, better antennas, higher mounting, or shorter distance.
Every country regulates radio transmissions, and alternative network builders must operate within these regulations. Ignorance is not a defense — illegal transmissions can result in fines, equipment seizure, and interference with critical services.
Key regulatory concepts:
EIRP (Effective Isotropic Radiated Power): the total radiated power, combining transmitter power and antenna gain. Regulations typically limit EIRP rather than transmitter power alone. In the US, the FCC limits EIRP to 36 dBm (4 watts) for point-to-multipoint and higher for point-to-point links in the 5 GHz band, with complex rules depending on the specific band and use case.
Duty cycle: in some bands (especially in Europe), regulations limit the percentage of time a device can transmit. LoRa in the 868 MHz band in Europe is limited to 1% duty cycle in some sub-bands — you can transmit for at most 36 seconds per hour. This fundamentally affects protocol design and network capacity.
Channel bandwidth: regulations may limit the bandwidth of individual transmissions
Spurious emissions: regulations limit out-of-band emissions that might interfere with other services
For the ISM bands used by most alternative networks, the general rules are:
Ham radio frequencies offer much higher power limits (up to 1,500 watts in the US) and access to additional spectrum, but require a license and prohibit encryption and commercial use in most jurisdictions. For alternative networks that can operate within these constraints, ham radio opens up powerful long-range capabilities that ISM bands cannot match. We will explore this in detail in Chapter 9.
One practical note: antenna upgrades are often more effective than power increases. Doubling your transmit power gives you a 3 dB gain. Upgrading from a 5 dBi omnidirectional antenna to a 20 dBi directional antenna gives you a 15 dB gain — equivalent to increasing your transmit power by a factor of 32, while remaining well within regulatory limits. This is why antenna selection and placement are among the most important skills for alternative network builders.
We have covered a lot of ground in this chapter — protocol models, topologies, addressing, routing, performance characteristics, and wireless fundamentals. Let us tie it together with a concrete example of how these concepts apply to a real alternative network.
Imagine you are building a community mesh network for a rural valley. Here is how each concept from this chapter comes into play:
Topology: you choose a partial mesh. Houses within line-of-sight form mesh clusters using omnidirectional antennas. Mesh clusters connect to each other via directional point-to-point links across the valley. One node has a satellite internet connection (Starlink) that serves as the gateway. The result is a hybrid topology: mesh at the neighborhood level, point-to-point links as a backbone, with a star element at the gateway.
Physical layer: you use 5 GHz Wi-Fi for the backbone point-to-point links (high bandwidth, good for directional links) and 2.4 GHz for the mesh nodes (better range, better obstacle penetration). You add a Meshtastic LoRa mesh layer as a backup communication channel for text messaging when the Wi-Fi mesh is down.
Data link: you use BATMAN-adv for the Wi-Fi mesh, providing a flat Layer 2 network where all nodes appear to be on the same subnet.
Addressing: you use the 10.0.0.0/8 private IPv4 space, with different /24 subnets for each mesh cluster. IPv6 link-local addresses provide automatic configuration. You run dnsmasq on a Raspberry Pi for local DNS, resolving name.valley.local addresses.
Routing: BATMAN-adv handles mesh routing at Layer 2. For the backbone links, you run Babel to find optimal paths between mesh clusters. If the directional link between two clusters fails, Babel automatically routes through an intermediate cluster.
Performance expectations: the backbone links achieve 100–200 Mbps. Within a mesh cluster, throughput ranges from 10–50 Mbps depending on hop count. LoRa backup provides 1–10 kbps for text messaging. Latency within the mesh is 5–30 ms; through the Starlink gateway to the internet, 50–100 ms. Not gigabit fiber, but more than adequate for web browsing, email, video calls, and local file sharing.
Wireless planning: you survey the valley with binoculars and topographic maps. Backbone links need clear Fresnel zones, so antennas go on rooftops or short poles. Mesh cluster nodes need omnidirectional coverage, so they are mounted at gutter height. You calculate link budgets for the longest point-to-point links and confirm they have adequate margin. You keep all transmit powers within your country’s ISM band regulations.
This is a realistic scenario. Networks like this exist today, built by volunteers with modest budgets. The fundamentals in this chapter are what make it possible to design, deploy, and troubleshoot such a network with confidence.
In the next chapter, we will zoom in on the building blocks: local area networks, the hardware that runs them, and the services that make them useful even without any internet connection.
| ← Previous: Introduction to Alternative Networks | Table of Contents | Next: Local Area Networks → |