Daniel sent through a two-parter today that started with a photograph of the British consulate here in Jerusalem, and that very tall antenna they've got mounted up top. We were talking earlier about why a diplomatic mission bothers with HF at all when the whole world runs on fiber and cellular. The answer we kept landing on is that HF doesn't depend on anything the host government owns. A cellular network is base stations and fiber and switching centers, all of which can be switched off by whoever controls the territory. The ionosphere is not a switch anyone can flip. But then Daniel pushed on the physics. He wants to know whether the lossiness of HF actually scales with transmission length, because it seems intuitive that it would, and he's right to test that assumption. And then he asked the more interesting question. A global diplomatic network is, topologically, a worldwide distributed nodal network. So are there mesh networking approaches that work at that intercontinental scale, where each embassy relays traffic for its neighbors rather than everyone just shouting at a head office? And are there documented systems where that nodal approach has actually been deployed at that scale?
So let's start with the physics, because the answer to Daniel's first question is not what your intuition says, and the reason it's not is the whole episode in miniature.
The antenna on the consulate is the give away. That thing is not pointed at a satellite. It's pointed at the sky in a way that says the building underneath it does not trust the local phone company.
Right. HF, three to thirty megahertz, propagates by skywave. The signal goes up, refracts off the ionosphere, comes back down hundreds or thousands of kilometers away. That's why it reaches beyond line of sight with no infrastructure in between. But that same ionospheric dependence is what makes it variable. Day and night change the ionosphere's density. Solar activity changes it. Geomagnetic storms change it. So the effective range and the reliability of a given frequency are moving targets.
Which is why the consulate has an antenna that looks like it was built to survive a siege. Because the medium it's using is the least governable thing in the entire stack.
And here's where Daniel's first question gets interesting. The intuitive model is free-space propagation. Signal spreads out as it travels, so loss scales with distance. That's true for line-of-sight microwave, for cellular, for satellite uplinks. But HF skywave is not free space. The signal bounces off the ionosphere, sometimes multiple times. Each bounce is a hop. And each hop has its own loss characteristics that depend on the frequency, the ionospheric conditions, the takeoff angle, and how many times the signal has already bounced.
So loss is a function of the number of bounces, not the number of kilometers.
A single hop under good conditions can cover roughly three to four thousand kilometers. Beyond that, you need multi-hop propagation. The signal bounces between the ionosphere and the ground, and each bounce introduces additional loss. But the relationship is not linear with distance. A ten thousand kilometer path might use three hops and have less total loss than a five thousand kilometer path that requires a high-angle, weak reflection.
Wait. So the shorter path can be the worse path.
Yes. That's the part that breaks the intuition. If the ionosphere is not cooperating at the frequency you're using, a short path can have terrible loss. Meanwhile a long path that catches good reflections on every hop can be surprisingly clean. The distance is almost secondary. What matters is the hop count and the quality of each reflection.
So when Daniel asks whether lossiness scales with transmission length, the answer is no, not in the way he means. It scales with the number of hops and the state of the ionosphere along the way.
And there's a formal way to think about this. Any given frequency has a maximum usable frequency, the MUF, above which the ionosphere won't refract it back down. It just punches through into space. And there's a lowest usable frequency, the LUF, below which the signal gets absorbed before it can bounce. The usable window between those two moves with time of day, season, and where we are in the solar cycle. So the operator is always choosing a frequency that fits the current conditions.
Which is why HF operators talk about the ionosphere the way sailors talk about weather. It's not a fixed medium. It's a condition you read and adapt to.
And that variability is the whole problem for running IP over HF. The modems can do it. The standards exist. MIL-STD-188-110C, STANAG 4539. Data rates from seventy-five bits per second up to ninety-six hundred bits per second or higher under good conditions. But the bit error rate is high and it swings around. And TCP/IP assumes a relatively stable path with low loss. When TCP sees packet loss, it assumes congestion and backs off. Over HF, that's catastrophic. The loss isn't congestion. It's the ionosphere being the ionosphere. But TCP doesn't know that.
So the protocol is fighting the medium.
Constantly. The fix is either a performance-enhancing proxy that sits between the TCP endpoints and pretends the link is healthier than it is, or you abandon TCP's assumptions entirely and move to delay-tolerant networking. The Bundle Protocol, RFC 5050. Store and forward. A node receives a bundle, holds it, and forwards it when a link becomes available. That's how the Interplanetary Internet is designed to work, because Mars doesn't have a stable low-latency link either.
So the same architecture that handles a rover on Mars is the one that handles an embassy in a country where the local infrastructure just got switched off.
And that's the bridge to Daniel's second question. If HF links are variable and lossy, can you build a mesh network out of them? Because the diplomatic network is already a distributed nodal system. Embassies all over the world, each with HF capability. The question is whether they can relay for each other.
And the honest answer is that a true mesh at intercontinental scale is theoretically possible, but practically it runs into the same problem TCP does. The routing protocols that make mesh networks work assume links are fairly stable and latency is low.
OLSR, BATMAN, AODV. These are the classic mesh routing protocols. They work beautifully on community WiFi networks where links come and go but latency is measured in milliseconds and bandwidth is measured in megabits. Over HF, a link might be there for twenty minutes and gone for two hours. Latency is seconds at best. Bandwidth is a few kilobits per second. The routing table updates alone would flood the network.
So you need the store-and-forward approach. A node receives a bundle, holds it until a link opens up, then forwards it. That's DTN. That's the Bundle Protocol.
And here's the thing. DTN was designed for exactly this scenario. Intermittent connectivity, long delays, high loss. The research on interplanetary mesh routing is directly relevant, because it's solving the same problem. How do you route traffic when links are intermittent and delays are measured in minutes or hours, not milliseconds?
So the answer to Daniel's second question is yes in theory. The architecture exists. DTN can do intercontinental mesh. But has anyone actually deployed it as a full mesh for diplomatic traffic?
The documented systems are more hub-and-spoke than true mesh. The U.S. diplomatic HF network, which is covered in the Foreign Affairs Manual, 5 FAM 540, is about emergency and evacuation communications. Embassies have HF capability, but the topology is mostly point-to-point or star. An embassy talks to a regional hub or to a central station. It's not relaying traffic for its neighbors as a matter of routine.
Which makes sense when you think about the operational reality. A true mesh means every embassy is a router. That means every embassy needs the antenna, the power, the frequency coordination, and the trained operators to handle traffic for other embassies. That's a lot of overhead for a link that might carry a few kilobits per second.
And the routing tables would be enormous. If you have a couple hundred embassies, each with multiple HF frequencies, and the links between them are constantly changing with ionospheric conditions, the link-state updates alone would eat your entire bandwidth budget. You'd spend all your time updating the routing table and none of it moving actual traffic.
So the mesh is the fallback, not the primary architecture. It's what you fall back to when the hub is unreachable and you need to get a message out through whatever path exists.
And that's where DTN shines. Because DTN doesn't need a routing table in the traditional sense. It needs a forwarding decision at each hop. Given this bundle, and given the links I have right now, where do I send it next? That's a much simpler problem than maintaining a global topology map.
Isode has done work on this. They have a whitepaper on IP routing over HF, and another on supporting mobility over multiple HF networks. They're thinking about exactly this problem, how to make routing decisions when the links are intermittent and the nodes are moving.
The mobility piece is interesting because it's not just embassies. It's also military units, ships, aircraft. Anything that moves and needs to communicate over HF. The routing problem gets harder when the nodes themselves are changing position.
So what's the second-order implication here? If you did deploy a true global HF mesh, what would you actually get?
You'd get something that is extremely resilient to jamming and infrastructure attacks. Because there's no single point of failure. No central hub you can take out. No fiber you can cut. Every node can relay for every other node. The network would survive almost anything short of a solar storm that wipes out the ionosphere itself.
But it would be slow.
Painfully slow. A few kilobits per second on a good day. Delays of seconds to minutes, or longer if a node has to wait for a link to open up. That's fine for text messages and emergency traffic. It's not fine for the modern diplomatic workload of email with attachments, file transfers, video conferencing. So the mesh is a fallback, not a replacement.
Which brings us back to the consulate antenna. It's not there to replace the fiber. It's there to be the thing that still works when the fiber is gone.
And that's the misconception most people have about HF. They think it's obsolete because satellites exist. But satellites have ground stations. Ground stations are infrastructure. Infrastructure can be attacked or denied. HF's independence from any infrastructure, including satellite ground stations, is what makes it a permanent fallback.
The ionosphere doesn't have a ground station. It doesn't have an owner. It doesn't have a service level agreement.
It has a solar cycle. Which is its own kind of unpredictability. But it's not the kind of unpredictability that a hostile government can exploit by flipping a switch.
So Daniel's first question, does loss scale with distance, the answer is no, not in the way intuition says. It scales with hop count and ionospheric conditions. A short path can be worse than a long one.
And his second question, can mesh networking work at intercontinental scale, the answer is yes in theory, with DTN and the Bundle Protocol. But the documented deployments are more hub-and-spoke than full mesh. The mesh is the fallback architecture, not the operational one.
There's a detail in the physics I want to sit with for a second. The hop count thing. You said a single hop covers three to four thousand kilometers under good conditions. So a path from Jerusalem to Washington, that's what, nine and a half thousand kilometers? That's three hops minimum.
Roughly. And each hop is a chance for the signal to degrade. But the degradation isn't uniform. The first hop might be clean, the second might hit a disturbed region of the ionosphere, the third might be fine. The total loss is the sum of what happened at each hop, not a function of the total distance.
So when an HF operator is trying to reach a specific station, they're not just picking a frequency. They're picking a frequency that works for the specific hop structure of that path at that time of day.
And that's where the skill comes in. You can't just set a frequency and walk away. You have to understand the ionosphere, the solar conditions, the seasonal patterns. It's a craft as much as a science.
Which is why the consulate probably has someone whose entire job is knowing what frequency to use to reach London at three in the morning local time during a geomagnetic storm.
Or at least a very well-configured automated system. The modern HF modems do a lot of this automatically. They probe the channel, they negotiate data rates, they adapt. It's not the nineteen fifties anymore. But the physics hasn't changed.
The ionosphere is still the ionosphere. It doesn't care about your automation.
Right. And that's the thing Daniel's question gets at. The physical atmospheric constraint isn't going away. You can engineer around it, but you can't eliminate it.
Now, the mesh question. Let me push on something. Daniel framed it as each embassy serving as a relay for its neighbors. That's the mesh ideal. But what does that actually buy you in the diplomatic context?
Redundancy. If the embassy in Cairo can't reach Washington directly, it can route through Amman or Nairobi or wherever the ionosphere happens to be cooperative. The mesh gives you multiple paths, and DTN lets you use them opportunistically.
But the cost is that every embassy has to be listening for traffic that isn't its own. That's a real operational burden. You need antennas pointed in multiple directions, you need receivers running around the clock, you need people or systems to make forwarding decisions.
And you need frequency coordination. If every embassy is transmitting on HF, you have to make sure they're not interfering with each other. That's a global coordination problem. The ITU allocates HF frequencies, but within that allocation, a mesh of a couple hundred nodes all trying to use the same band is a recipe for interference.
So the mesh is theoretically elegant and operationally messy.
That's the short version. The long version is that DTN solves the routing problem, but it doesn't solve the spectrum problem or the antenna problem or the power problem or the personnel problem.
And those are the problems that actually determine whether a network gets deployed.
Which is why the documented systems are hub-and-spoke. It's simpler. One central station, many remote stations. The central station has the big antenna and the skilled operators. The remote stations have smaller setups. Traffic flows through the hub. If the hub is down, you have a problem, but the hub is usually in a friendly country with reliable infrastructure.
So the hub-and-spoke model is an infrastructure bet. You're betting that the hub stays up.
And the mesh is a no-infrastructure bet. You're betting that the network can survive even if any given node goes down. But you're paying for that resilience with complexity.
There's a nice parallel here with the internet itself. The early internet was designed to be a mesh. Every node could route for every other node. That's the whole point of packet switching. But the modern internet is much more hub-and-spoke, with a few giant content delivery networks and transit providers handling most of the traffic.
The economics pushed it that way. It's cheaper to centralize. But the original design principle is still there underneath. And in a crisis, that mesh capability is what lets the network route around damage.
So the diplomatic HF network is like the early internet. The mesh capability is there in principle, but the day-to-day operation is more centralized because that's what's practical.
And the question Daniel is really asking is whether the mesh capability can be made real at the scale of a global diplomatic network. And the answer is that the technology exists, but the operational reality pushes against it.
Let me ask you something. The Bundle Protocol, RFC 5050. How does it actually make a forwarding decision? If there's no global routing table, how does a node know where to send a bundle?
It uses a few different approaches. One is scheduled contacts. If you know that a link to a particular node will be available at a particular time, you hold the bundle until then and forward it. Another is opportunistic contacts. If a link opens up unexpectedly, you forward whatever bundles seem relevant. And there's a third approach called custody transfer. A node accepts custody of a bundle, which means it takes responsibility for getting it to the next hop. That way the sender knows the bundle isn't lost, it's being held.
It's less like routing and more like a postal system with relay stations.
That's actually a very good way to think about it. The Pony Express, or the postal relays in old empires. A message moves from station to station, each one holds it until the next leg is available, and eventually it reaches the destination. The routing decision is local at each station, not global.
Which is why it works for interplanetary communication. Mars is only in line of sight with Earth at certain times. You hold the bundle until the next contact window.
The same principle applies to HF. You hold the bundle until the ionosphere opens up a path to the next node.
The mesh isn't a real-time network. It's a store-and-forward network. The latency is measured in minutes or hours, not milliseconds.
Which is fine for a diplomatic cable that says we need evacuation assistance. It's not fine for a video call.
That's the tradeoff. Resilience against speed. The mesh gives you resilience, but it costs you speed.
For diplomatic emergency communications, resilience is the whole point. You're not trying to stream video. You're trying to get a message through when the normal channels are down.
Daniel's instinct is right. The mesh approach is the right architecture for the problem. It's just that the problem is narrower than the full diplomatic workload.
The mesh is for the emergency. The hub-and-spoke is for the routine. And the two coexist.
Let me ask you about the Isode work. What are they actually doing that's relevant here?
They've been working on HF networking for a long time. Their whitepaper on IP routing over HF is about how to make standard IP applications work over HF links. The key insight is that you need a routing protocol that understands the link is intermittent and lossy. Standard routing protocols like OSPF or BGP assume stable links. They'll flap and reconverge constantly over HF.
You need a routing protocol that's patient.
A routing protocol that doesn't panic when a link disappears for twenty minutes. That treats link loss as normal, not as a failure.
Their mobility work is about nodes that move. Ships, vehicles, aircraft. The routing problem gets harder when the nodes themselves are changing position.
Because the topology is changing for two reasons. The ionosphere is changing, and the nodes are moving. That's a doubly dynamic network.
Which is a nice way to describe the diplomatic network too, actually. The embassies don't move, but the ionosphere between them does.
The diplomatic network has the easier version of the problem. Fixed nodes, dynamic medium. The military has the harder version. Moving nodes and dynamic medium.
The military has been working on this for decades. Mobile ad hoc networking, tactical data links, all of that.
That's where a lot of the DTN research comes from. The military needs networks that work when everything is broken. That's the same requirement as the diplomatic emergency network.
The answer to Daniel's question about documented systems is that the military has deployed versions of this, but the diplomatic network itself is more conservative.
The diplomatic network has the capability, but it's not running a full mesh day to day. It's running hub-and-spoke with the mesh as a fallback.
Which is probably the right call, honestly. The full mesh is a lot of complexity for a capability you hope you never need.
But when you need it, you really need it. That's the insurance argument. You don't buy insurance because you expect the house to burn down. You buy it because if it does, you want to be covered.
The consulate antenna is the insurance policy made visible.
It's a very tall insurance policy.
Let me try to summarize what we've established. Daniel's first question, does HF loss scale with distance. The answer is no. It scales with hop count and ionospheric conditions. The intuition is wrong because it assumes free-space propagation, but HF is a waveguide. The ionosphere and the ground act as boundaries, and the loss depends on how well those boundaries reflect at your frequency.
That's the physics. And it's why HF operators talk about MUF and LUF and solar cycles instead of just distance.
Daniel's second question, can mesh networking work at intercontinental scale. The answer is yes in theory, with DTN and the Bundle Protocol. But the practical constraints, antenna size, frequency coordination, power, skilled operators, mean that most diplomatic HF networks are not full meshes. They're hybrid. Some point-to-point, some hub-and-spoke, with store-and-forward for emergencies.
The documented systems confirm that. The U.S. diplomatic HF network is about emergency communications, not routine mesh routing.
The nodal approach that Daniel is asking about exists in principle and in some military deployments, but the diplomatic network is more conservative.
Which is a reasonable engineering decision, even if it's not the most elegant topology.
Now, there's a second-order implication I want to explore. If a true global HF mesh were deployed, what would it mean for the threat model?
It would be extremely hard to take down. You'd have to jam every node simultaneously, or disrupt the ionosphere itself, which is not something any adversary can do. So the mesh is the ultimate resilience play.
But it's also slow. So an adversary doesn't need to take it down. They just need to make the normal channels unreliable, and force you onto the slow mesh. That's a degradation attack, not a denial attack.
That's a really interesting point. The mesh doesn't have to be taken down to be neutralized. If you can force the diplomatic network onto HF, you've already slowed it down by orders of magnitude. That's a win for the adversary even if the HF still works.
The HF mesh is a deterrent, but not a full counter. It says you can't silence us completely. But it doesn't say you can't slow us down.
Which is why the diplomatic network invests in satellite and fiber and everything else. HF is the last resort, not the primary.
That's the right frame for the consulate antenna. It's not the primary communications system. It's the thing that still works when everything else is gone.
The antenna is a promise that the mission can always get a message out, no matter what the host government does.
That's a very specific kind of insurance. And it's worth the cost of a very tall antenna.
The cost of someone who knows how to use it. Which is the part people forget. The antenna is just metal. The skill is in the operator.
The ionosphere doesn't care about your antenna budget. It cares about your frequency choice and your timing.
Your willingness to try again when the first attempt doesn't get through.
Which is a nice segue to something I wanted to ask you. The IP over HF problem. You said TCP backs off when it sees packet loss. But what actually happens in practice when someone tries to run a web browser over an HF link?
It's painful. The page load times out. The connection resets. The browser retries. The whole thing is a mess. That's why the performance-enhancing proxies exist. They sit at each end of the HF link and spoof the TCP acknowledgments, so the TCP endpoints think they're on a fast, reliable link. The proxy handles the retransmissions and the buffering.
The proxy is lying to the protocol.
In a sense, yes. It's telling TCP what TCP wants to hear, and then dealing with the messy reality underneath. It's a translation layer between the protocol's assumptions and the medium's actual behavior.
Which is a nice metaphor for a lot of engineering. You build a layer that lies to the layers above it so they don't have to know how bad the layers below are.
That's basically the entire internet. Every layer lies to the layer above it.
HF is just a particularly extreme case of the lying being necessary.
Because the medium is so far outside what the protocols were designed for. TCP was designed for a network with low loss and low latency. HF is high loss and high latency. The mismatch is fundamental.
The Bundle Protocol is the honest approach. It doesn't try to hide the medium's behavior. It accepts it and works with it.
Store and forward. Accept that the link will be intermittent. Accept that the delay will be long. Design for that reality instead of pretending it doesn't exist.
Which is why DTN is the right architecture for the mesh. It doesn't fight the medium. It works with it.
That's the deep insight in Daniel's question. He's asking whether the mesh can work at intercontinental scale. The answer is that it can, but only if you accept the medium's constraints and design for them. You can't just take a WiFi mesh protocol and run it over HF. You need a different architecture.
The architecture exists. It's DTN. The question is whether anyone has deployed it at that scale for diplomatic traffic.
The answer is mostly no. The deployments are smaller, or they're military, or they're experimental. The diplomatic network is more conservative.
Daniel's question identifies a real gap between what's theoretically possible and what's actually deployed.
Which is a very Daniel thing to ask. He's always looking for the gap between theory and practice.
The gap is where the interesting problems live.
Where the interesting engineering happens.
If someone wanted to close that gap, what would they need to solve?
The frequency coordination problem is the big one. A global mesh of HF nodes needs a way to share spectrum without interfering. That's not a solved problem for a network of that scale.
The routing problem. DTN solves the store-and-forward, but the forwarding decisions at each node are still hard. You need to know which neighbor is most likely to have a path to the destination.
That's the research frontier. The interplanetary DTN work is addressing exactly that. How do you make forwarding decisions when you have incomplete information about the network state?
The answer to Daniel's question is that the approach exists, the architecture exists, but the deployment at intercontinental scale for diplomatic traffic is still more aspiration than reality.
The reason is not that it's impossible. It's that the operational costs are high and the benefit over hub-and-spoke is only realized in the emergency case.
Which is the insurance argument again. You pay for the mesh capability even if you hope you never need it.
The question is whether the premium is worth it. For a diplomatic network, probably yes, because the emergency case is exactly when you can't afford to be offline.
The consulate antenna is a bet that the emergency case is real enough to justify the cost.
Given the world we live in, that's not a hard bet to justify.
No, it's not.
Hilbert: I worked for a company that made HF antennas for maritime use. This was the late eighties. I spent six months on a ship in the North Atlantic testing a new whip antenna. The thing kept icing up. I had to climb the mast in a storm to chip ice off the loading coil.
You climbed the mast in a storm.
Hilbert: I was younger and dumber. The point is, HF is not a set it and forget it technology. You have to tune it. You have to know the ionosphere. You have to be willing to climb the mast. The diplomatic network is probably full of people who know this. That's why it still works.
The physicality of it. It's not just software. It's antennas and weather and human beings doing maintenance.
Hilbert: I still have my ham radio license. Call sign W2XYZ. I once talked to a station in Antarctica on twenty meters with a hundred watts. The ionosphere doesn't care about your politics or your embassy. It just reflects.
A hundred watts to Antarctica. That's the waveguide thing Herman was describing. The signal bouncing between the ionosphere and the ground all the way down there.
Hilbert: It was three in the morning. The band was open. I had a wire antenna strung between two trees. Nothing fancy.
That's the thing about HF. You don't need a massive antenna to reach across the planet. You need the right conditions and the right frequency.
Hilbert: The willingness to be awake at three in the morning when the band opens.
The consulate antenna is not about raw power. It's about having the right tool for when the conditions align.
Hilbert: Someone who knows how to use it. The antenna is just metal. The operator is the system.
The operator is the system.
Hilbert: The company I worked for went under in ninety-two. The maritime HF market collapsed when satellite phones got cheap. But the antennas we built are probably still on ships somewhere. They don't break. They just ice up.
Someone has to climb the mast to chip the ice off.
Hilbert: That's the job. That was always the job.
The misconception most people have about HF is that it's obsolete. But the reality is that it's the one communications medium that doesn't depend on any infrastructure you don't control.
The one that still works when everything else is switched off.
Which is why the consulate has that very tall antenna. Not because it's the best option. Because it's the only option that can't be taken away.
The open question is whether the mesh approach will ever be deployed at the scale Daniel is asking about. The technology exists. The architecture exists. But the operational costs are high, and the hub-and-spoke model is good enough for the routine case.
As satellite constellations expand, the pressure to invest in HF mesh decreases. Why build a complex HF mesh when you can buy a Starlink terminal?
But Starlink has ground stations. And ground stations are infrastructure. And infrastructure can be denied.
HF remains the fallback. Not because it's fast. Because it's independent.
That independence is worth a very tall antenna and someone who knows how to use it.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop. If you have a weird prompt, send it to us at show at my weird prompts dot com. We'll be back soon.
