Internet Protocol Over Avian Carriers
Internet Protocol Over Avian Carriers (IPoAC), or, as I prefer, Pigeon Over Internet Protocol (PoIP), may sound like a bad joke. And while it is humorous engineering, it is also a formally specified and technically valid data transportation method, as described in RFC 1149. In certain niche scenarios, PoIP can even outperform modern communication systems in throughput-bound situations, while serving as an unconventional, low-tech redundancy mechanism. This post takes a look at the satirical protocol, dissects the real engineering behind it, and makes some deeply unfair comparisons with modern networks.
Pack your data and tighten your feathers.
Summary of IPoAC
Long before fiber optics, radio, or satellites, avians, most notably pigeons, played a central role in long-distance communication. They were faster than horses, slower than light, and offered limited bandwidth, as sadly, you cannot ask a pigeon to carry a book. Today, we rely on fiber, copper, RF links, and satellites for data transport. Unless you’re a wizard. This article, however, is grounded strictly in real engineering. No magic allowed.
IPoAC is formally defined in RFC 1149, later clarified by RFC 2549, which specifies how IP packets may be encapsulated in physical payloads transported by birds. For this post, we’ll focus on pigeons as the carrier bird, since they’ve been used throughout history, are trainable, reliable over distance, and consistent in return paths. Other species could be valid carriers as well and would bring new trade-offs to the protocol. Eagles would offer high payload capacity but impractical logistics, hawks could bring the best latency but are not the most cooperative couriers, and owls might provide excellent night ops with questionable throughput metrics.
While written as an April Fools’ RFC, it accurately models a delay-tolerant, high-bandwidth, extremely high-latency network. In modern terms, PoIP fits neatly into the Delay-Tolerant Networking (DTN) category.
Key characteristics:
- Medium: Physical transport (avian carrier)
- Latency: Minutes to hours (sometimes worse)
- Throughput: Potentially enormous
- Packet loss: Non-zero and occasionally catastrophic
- Security: Naturally air-gapped (ISO 27001 auditors hate this one simple trick)
Latency & Throughput
As can be expected, latency for PoIP is absolutely terrible: minutes to hours, sometimes days. Homing pigeons fly at an average of 55 km/h, maybe reaching 90 km/h in brief periods, and these values are highly dependent on weather conditions, much like satellite communications. So, sending data between two nearby cities means waiting at least around a full hour.
Throughput, however, is surprisingly high. A good USB-C pendrive can easily sustain ~1000 MB/s read and 800–900 MB/s write throughput. Some SD Express memory cards can reach ~1600 MB/s read and ~1200 MB/s write speeds.
Once airborne, storage performance becomes irrelevant. What matters is payload size divided by delivery time.
Assumptions
- Average pigeon speed: 55 km/h
- Speed in meters per second:
- Payload: 1 TB SD Express card
1 TB ≈ $10^{12}$ bytes ≈ 8 terabits - Operator delay is ignored for lack of data and practicality
Maximum Transmission Unit
Every network needs an MTU, and PoIP is no exception, even if defining one feels slightly ridiculous.
In traditional networks, the MTU is constrained by the physical link: Ethernet frames, PPP encapsulation, radio limitations, and so on. In PoIP, the “link” is the avian carrier itself, combined with whatever storage medium it is transporting.
For the sake of practicality and realism, we define the PoIP MTU as:
The maximum IP payload that can be transported by a single avian carrier.
In other words, the MTU is bounded by the capacity of the attached storage medium, minus whatever minimal encapsulation, filesystem, and metadata overhead is required. If a pigeon carries a 1 TB SD Express card, then the PoIP MTU is effectively ~1 TB.
Yes, this is a jumbo MTU. So large, it makes Ethernet cry.
Fragmentation and Reassembly
Fragmentation in PoIP does not happen at the byte or frame level. Instead, it happens at the carrier level.
- A dataset larger than the MTU is fragmented across multiple pigeons.
- Each carrier transports a self-contained chunk of the payload.
- Reassembly occurs at the destination once all carriers arrive (or enough arrive, depending on redundancy).
This is entirely consistent with delay-tolerant, store-and-forward networking. Losing a carrier is equivalent to losing a very large packet.
Operational Considerations
While the theoretical MTU is limited only by storage capacity, a real PoIP operator would likely impose a lower operational MTU to limit blast radius:
- Losing one pigeon should not mean losing all data.
- Smaller chunks improve resilience at the cost of additional carriers.
- Redundancy, erasure coding, or replication can be layered on top.
As with most networking decisions, this becomes a trade-off between efficiency, reliability, and how emotionally invested you are in individual packets.
In summary, PoIP supports MTUs of unprecedented size, proving that nothing in IP explicitly forbids absurdity, as long as the packets eventually arrive.
Latency & Effective Bandwidth Calculations
1 km Trip
Time: \(t = \frac{1000}{15.28} \approx 65.4\ \text{s}\)
Effective bandwidth: \(B = \frac{8\ \text{Tb}}{65.4\ \text{s}} \approx 122\ \text{Gb/s}\)
≈ 15.3 GB/s
This comfortably outperforms:
- 100 Gb Ethernet
- USB4
- Thunderbolt 4
(While being catastrophically worse on ping.)
10 km Trip
Time: \(t = \frac{10{,}000}{15.28} \approx 654\ \text{s} \ (\approx 10.9\ \text{min})\)
Effective bandwidth: \(B = \frac{8\ \text{Tb}}{654\ \text{s}} \approx 12.2\ \text{Gb/s}\)
≈ 1.53 GB/s
Comparable to fast NVMe-over-USB or entry-level datacenter links.
1000 km Trip
Time: \(t = \frac{1{,}000{,}000}{15.28} \approx 65{,}400\ \text{s} \ (\approx 18.2\ \text{h})\)
Effective bandwidth: \(B = \frac{8\ \text{Tb}}{65{,}400\ \text{s}} \approx 0.122\ \text{Gb/s}\)
≈ 15.3 MB/s
At this point, PoIP degrades to fast DSL / slow Ethernet, with latency that would send TCP into an existential crisis.
Average Numbers (Pipeline Scenario)
If you can launch one pigeon every minute, each carrying 1 TB:
\[B = \frac{1\ \text{TB}}{60\ \text{s}} \approx 16.7\ \text{GB/s}\]At this scale, the limiting factors are:
- pigeon handling
- loading and unloading
- pigeon air traffic control
- seed logistics
In short, latency is awful, but bandwidth is hilariously high.
Data Security
Data security and integrity are another aspect where PoIP really shines. Encryption at rest is absolutely mandatory, but with modern full-disk encryption using AES-256 (XTS) or ChaCha20-Poly1305, paired with a strong passphrase, data is effectively secure in case of carrier interception. The passphrase must be provisioned beforehand and must not travel with the carrier itself.
Additionally, PoIP is secure by physical limitations:
- No exposed ports
- No RF emissions
- Totally immune to remote exploitation
PoIP is naturally air-gapped. Data may fly slowly, but it flies safely. From the data’s point of view, at least.
Threat Model
While the data itself is safe and virtually impossible to exfiltrate remotely, the carrier faces risks unknown to traditional protocols. Packet loss may be caused by:
- weather conditions (routing instability)
- fatigue
- navigational errors
- natural predators (hard packet drops)
PoIP is also vulnerable to physical DoS attacks, though attackers must be present along the route and have excellent aim. To be fair, modern DDoS attacks are still cheaper and easier against conventional networks.
Recovery is straightforward, though it may involve potentially heavy emotional costs if operators are sentimentally attached to their carriers.
Privacy
PoIP offers exceptional privacy. With no RF emissions, telemetry, or logs, correlating traffic requires physical observation of the carrier. Large-scale surveillance becomes hilariously impractical and deeply inefficient.
Advanced setups may include GPS tags or logging, but these require deliberate manual effort.
Comparisons (With Numbers, Because Engineering)
PoIP vs Starlink
Starlink (clear sky):
- Downlink: 100–250 Mbps
- Uplink: 10–40 Mbps
- Latency: 25–50 ms
Starlink (heavy rain):
- Throughput drops below 20–50 Mbps
- Temporary link loss due to rain fade
- Latency spikes during handovers
For short distances (≤10 km) and bulk payloads (≥1 TB), PoIP achieves >1 GB/s effective throughput, at the cost of minutes of latency. If RF conditions are hostile and latency is irrelevant, the pigeon wins.
PoIP vs Underwater Fiber
- Transatlantic fiber:
- Aggregate capacity: 200–400 Tbps
- Single wavelength: 100–400 Gbps
- Latency: ~60–80 ms
Fiber dominates:
- latency
- reliability
- sustained throughput
But when a submarine cable fails:
- repairs take days to weeks
- require ships, permits, and geopolitics
PoIP restoration time: minutes, assuming spare pigeons.
PoIP vs Wi-Fi
Wi-Fi:
- Range: 20–100 m
- Throughput: hundreds of Mbps
- Attack surface: RF sniffing, deauth, KRACK-class issues
PoIP:
- Range: global
- RF emissions: none
- Remote exploitation: physically impossible
Wi-Fi wins for interactivity.
PoIP wins for air-gapping, range, and attack surface minimization.
User Sweet Spot
PoIP excels at large, non-interactive, delay-tolerant data:
- backups
- cold archives
- scientific datasets
- AI models
- regulatory data dumps
- blog backups (this one included)
If latency doesn’t matter and throughput does, PoIP starts to look… uncomfortably reasonable.
Redundancy System
Despite the humor, PoIP is a legitimate redundancy option during:
- power outages
- network blackouts
- EM interference
- infrastructure collapse
- extreme surveillance environments
Examples include off-site backups when all digital communication is unavailable, or physically exporting critical data from an isolated zone in a sensitive time frame.
Conclusion
Pigeons have been an integral part of long-range human communications for more than 5,000 years, even carrying encrypted messages long before radio flooded the ether. Applying them to digital communication is humorously entertaining, but not entirely absurd. If birds could carry messages in the age of paper, they can surely carry large amounts of data in the era of digital systems and data compression.
PoIP is not practical for interactive workloads, real-time systems, or anything involving user patience. It will not allow you to play League of Legends or Counter-Strike unless you enjoy one frame every several minutes or hours, but it could deliver your Netflix content without buffering problems. With enough patience, it even allows you to play a game of Civilization with a same-city opponent, in extremely slow, true marathon-style turns.
As a thought experiment grounded in real engineering, PoIP highlights an important truth: bandwidth and latency are orthogonal problems. They measure fundamentally different properties of a network. Improving one does not imply improving the other, and systems optimized for bulk data transfer often perform poorly for interactive traffic.
While practicality will remain ultra-niche, PoIP is a valid consideration when moving large amounts of data where digital communications are impossible, particularly during disasters, in cities experiencing communication blackouts or highly surveilled and privacy-violating networks. In those rare scenarios, it becomes a creative, if unconventional, solution.
And it reminds us that creative engineering, no matter how fun or silly, can still be applied to real problem-solving situations when based on real data. Sometimes, even if very rarely, the best network might just be the one with feathers.