I'm Getting Into Mesh Networks... (Meshtastic, MeshCore, and Reticulum)

I love networking, a lot. So much so that I've run my own ISP since 2024, complete with its own ASN, IPv4/6 address space, fiber optics, etc.However, do this and you quickly realize how reliant you still remain on central service providers. The internet is a mesh, but the real players on that mesh are few, far between, and easy to coerce into censorship and other bad things. Even after ascending to the lofty realms of direct BGP peering myself, my access to those resources is locked behind yearly fees from ARIN. Ownership of the "real estate" of the internet, IP addresses, no longer exists.The tragedy of modern computing is that the local compute we own in our offices, on our laps, or in the palm of our hands is massively, massively powerful, but Big Tech companies actively refuse to take advantage of that fact. Why are you, I, and our neighbors largely relegated to consuming access from big players when the computers we have are capable of so much more?Mesh networking, sending packets of data through many directly interconnected peers instead of through central datacenters, promises to free us from our reliance on central service providers, and it's something I've been really excited about lately.Of course, there is good reason for how the internet is currently designed. High bandwidth connections are costly, and for some applications lowering latency as much as possible is very important, which realistically requires continent- and ocean-spanning fiber optics with as few middle-men as possible.This doesn't mean we need to put all our eggs in this basket though. While bandwidth-intensive services like Netflix or latency-sensitive services like gaming are not likely to come to mesh networks anytime soon, there are a vast number of applications which are perfectly suited to mesh networking:Messaging, social networking, and general information sharing are very practical uses for mesh networking where access, censorship resistance, and resiliency are increasingly critical for many people around the world.For applications like this, we don't need to trench fiber connections through the ground to get everyone connected. In the modern mesh networking space, much of the innovation is happening in the LoRa radio space.LoRa radios use license-free, sub-gigahertz radio bands that are available for public use in nearly all countries around the world.

Compared to the license-free 2.4 GHz or 5 GHz radio you'd recognize from Wi-Fi (but is also used by many other technologies), LoRa operates at much lower power and, importantly and simultaneously, at much longer range.Mesh networking over the airwaves presents a very unique opportunity for our societies. We could build a resilient, peer-to-peer network that coexists with the internet, enabling connectivity in currently underserved regions and increasing our personal sovereignty online by maintaining a functional backup to the internet for our most critical needs.It's also just a freeing feeling to be able to send a message to someone else relying only on devices that you and people in your network own outright, instead of renting the capability to do so from your local ISP or Elon Musk's Starlink.MeshtasticThe obvious frontrunner in the mesh networking space is Meshtastic, mostly because they were the first in the consumer LoRa mesh space, or at the very least the first to do it pretty well.It's easy to see why Meshtastic has quite a bit of popularity: it's a real product designed with a specific use-case in mind (mobile messaging and device tracking, primarily), not a technical project just trying to build a network and hoping the use-case comes later.This is very appealing for most people who just want something they can buy and use out of the box, like a set of walkie-talkies from Wal-Mart. Unfortunately, much like those cheap walkie-talkies in comparison to more serious technology like amateur radio, Meshtastic's core design holds the platform back from achieving its full potential.Meshtastic's first-mover advantage is pretty hard to overcome, especially when it already works reasonably well for small, private groups like hikers or event-goers. For a very large and public mesh, however, it's become clear to most people that Meshtastic by design is a fairly untenable solution.

Some public mesh groups have increased the bandwidth available to Meshtastic by giving up some range, but it's a stop-gap solution that doesn't fix the problems Meshtastic has in this environment at the end of the day.💡To be perfectly upfront with you, this post will be glossing over many Meshtastic and MeshCore features, because I feel they are both non-serious solutions compared to Reticulum for reasons I will explain later on in this post. I can almost guarantee I have been running Meshtastic and MeshCore for longer and with more infrastructure than you, and in fact I still do, despite not really believing in their long-term success, so... Any omissions or "lack of problem solving" in this post are not due to a lack of knowledge, but simply because fully detailing the many problems they have is outside the scope of this article.I think most people who are serious about public mesh networking have moved on to researching other solutions, or will have to soon.MeshCoreMeshCore is one of these potential solutions some public mesh groups have begun switching to.While Meshtastic's original design essentially floods the network with every message being sent, hoping it eventually reaches the correct destination, MeshCore has an actual routing system that can send messages only through a path of specific devices on the mesh that include the sender and recipient.This results in a monumental reduction in radio transmissions, the advantages of which can't be understated. It makes the network less congested and more reliable, and for people who are mainly interested in messaging as opposed to sharing sensor/location data that Meshtastic remains well-suited for, it's no surprise that many larger groups have begun to shift to MeshCore.Unfortunately, MeshCore is not a true mesh in the way public mesh enthusiasts would probably like it to be. At a very high level, devices in MeshCore are broken up into two categories: companions and repeaters.Companion devices would be what most people use to send and receive messages, while repeaters are the devices which actually mesh with each other and extend the overall network's range. What this means is that a companion always has to be in range of a repeater to access the network, companions never relay messages between themselves on behalf of other companions.

There are advantages to this approach. MeshCore allows messages to traverse up to 64 hops away, which is an enormous real-world scale when LoRa repeaters can be many miles apart in ideal conditions. Even in the best possible case, Meshtastic's default 3-hop limit (albeit configurable up to 7) places a real limit on how far messages can spread.It's very true that anyone can participate in MeshCore as a repeater, so all the tools to build your own mesh are certainly there. It's just that it requires some additional planning, coordination, and centralization that I don't view as totally necessary.The bigger problem I have with MeshCore is that many parts of it are proprietary. While the underlying protocol and the firmware for some radios is open source, all of the official MeshCore clients are proprietary, and even have features paywalled. Proprietary software is simply not disaster-ready, and that goes doubly so for software reliant on central payment processors.I am not the type of person who needs to use open-source software 100% of the time by any means (although it'd be nice), but the only point in my mind of having an off-grid mesh network in the first place is total freedom and control, so in this particular case I simply cannot ever support a closed-source solution.Efforts are already being made to create unofficial open source clients for MeshCore. I won't discount this fact, but at the end of the day most people in the MeshCore ecosystem will be in the official, proprietary ecosystem, and I don't think MeshCore has enough advantages, users, or reliability to warrant adopting at this very early stage of mesh networking.We have a unique opportunity as enthusiasts to adopt the best mesh networking solution we can, before the "network effect" truly sets in and locks people in place to a particular platform. I think we can do better than Meshtastic and MeshCore.ProblemsUnfortunately, both Meshtastic and MeshCore are highly limited, and don't scale well. Meshtastic barely can scale to a regional mesh in ideal scenarios, and while MeshCore fares better here, it's still unlikely to scale to the size of many larger regions, much less countries or planets.The thing is, Meshtastic and MeshCore are both more applications than they are protocols.

They enable simple instant messaging communication over LoRa, but they don't give much thought to mesh networking applications beyond what their client apps officially support.They're geared towards communicating with a small, local group, and any public meshes on these networks are really exceptions rather than the standard use-case.Another problem is that Meshtastic and MeshCore both rely on LoRa pretty much exclusively.LoRa is very cool for building ad-hoc, low-bandwidth mesh networks, and it's a blessing that we have it available in most countries as an unlicensed option we can operate on without a ham radio license, and with modern digital technologies like encryption that are mostly forbidden on amateur radio.It's hardly the perfect solution for many scenarios though, and it is quite slow.In a perfect world, the mesh networking/routing software would be completely independent of the physical network that connects all these devices together. For example, I want to be able to build cheap, local LoRa networks in neighborhoods and other regional communities, and interconnect them with more powerful point-to-point microwave connections, or even fiber or the internet.Meshtastic (and I believe MeshCore via some unofficial "gateways") have some hacky methods of interconnecting different meshes with MQTT, but the experience is quite poor and it's clear that this type of connectivity is not a first-class experience on the network. Especially on Meshtastic, bridging to the internet with MQTT can degrade the network so much that it becomes impractical to use among any more than a handful of people.It should be possible to build a mesh routing solution that intelligently routes packets between nodes over many different types of connections, completely seamlessly so that the experience of using the mesh never changes based on your specific interface.Luckily for us, this has already been done!ReticulumReticulum is extremely cool.It's a networking stack that intelligently provides strongly-encrypted routing over a wide variety of physical networks, including LoRa. Like MeshCore, it has automatic routing over paths on the network, but unlike MeshCore, those paths can traverse not only over LoRa but over any supported interface.Additionally, like Meshtastic, Reticulum essentially works out of the box with devices operating on the same local network. Connect two devices on the same LoRa frequency and you have a functional mesh right away, with no advanced networking skills or dedicated repeaters required.