Meshtastic isn't very good (at least before their 2.6 release) [3].
It's clear they didn't research any historical mesh network schemes (ALOHAnet [0] and other MANETs [1]) when writing it. And flood routing more or less kinda worked, but as it go popular, it stopped working reliably. There's a video from Jeff Geerling about it and he was generous I think when he called Meshtastic "Beta" [4].
Meshtastic a few years ago released a youtube video describing how it worked, and there wasn't anything about topology resolution, it was pretty much about signal strength and device type [2].
This caused an issue because anyone could create a router. And often they did. And when they did, this could break routing because the router is in the other direction of where the originator wanted the traffic could go.
They also prioritized letting the edges forward the message on. AFAICT they could only detect this by signal strength. So a badly performing node (bad antenna or maybe a node turned on in a basement) could get priority.
The last issue is congestion. Nodes can send telemetry, often rather quickly, but that could get flooded on the network. And with a hop count max of 7, it often will go where it's not wanted, wasting the network bandwidth -- as nobody really cares about one particular node's battery life.
So in a dense Meshtastic metro (I can see multiple sites) I couldn't reliably get a message to a friend in the same city. The lesson is that the hardware is better than the software at this point. And there's no use using it until they fix their software. [3]
[3] I did some research on meshtastic and after finding and watching [2] I gave up on meshtastic, because it's clear that weren't super serious on routing algorithms, nor basic wikipedia reading. Version 2.6 maybe better, but there's a slew of nodes on 2.4 yet. And I don't want to bother with it anymore -- at least until they fix their reliability problems.
It's clear they didn't research any historical mesh network schemes (ALOHAnet [0] and other MANETs [1]) when writing it. And flood routing more or less kinda worked, but as it go popular, it stopped working reliably. There's a video from Jeff Geerling about it and he was generous I think when he called Meshtastic "Beta" [4].
Meshtastic a few years ago released a youtube video describing how it worked, and there wasn't anything about topology resolution, it was pretty much about signal strength and device type [2].
This caused an issue because anyone could create a router. And often they did. And when they did, this could break routing because the router is in the other direction of where the originator wanted the traffic could go.
They also prioritized letting the edges forward the message on. AFAICT they could only detect this by signal strength. So a badly performing node (bad antenna or maybe a node turned on in a basement) could get priority.
The last issue is congestion. Nodes can send telemetry, often rather quickly, but that could get flooded on the network. And with a hop count max of 7, it often will go where it's not wanted, wasting the network bandwidth -- as nobody really cares about one particular node's battery life.
So in a dense Meshtastic metro (I can see multiple sites) I couldn't reliably get a message to a friend in the same city. The lesson is that the hardware is better than the software at this point. And there's no use using it until they fix their software. [3]
[0] https://en.wikipedia.org/wiki/ALOHAnet
[1] https://en.wikipedia.org/wiki/Wireless_ad_hoc_network
[2] https://www.youtube.com/watch?v=7v6UbC5blJU
[3] I did some research on meshtastic and after finding and watching [2] I gave up on meshtastic, because it's clear that weren't super serious on routing algorithms, nor basic wikipedia reading. Version 2.6 maybe better, but there's a slew of nodes on 2.4 yet. And I don't want to bother with it anymore -- at least until they fix their reliability problems.
[4] https://www.youtube.com/watch?v=0A7A-CSd3e4