4 ms·

So you mean other than these sections right? "Thought experiments about mesh networking" "Hard Lessons Learned -- What not to do" "Meshtastic Is Rediscoverin

by bb88 9mo ago

So you mean other than these sections right? "Thought experiments about mesh networking" "Hard Lessons Learned -- What not to do" "Meshtastic Is Rediscovering Lessons (Already Learned) of Amateur Radio Data Networking" Instead of actually trying to understand the arguments these days, it's easier to inject noise into the argument, proclaiming it's too "hard to find" or "too hard to understand." Mesh networking is a hard topic. Expect to expend some brain cells to understand it. I'm not here to spoon feed you tech that was well understood 3 decades ago.

wtallis 9mo ago
How about you make an actual argument here in this thread, instead of vaguely gesturing at an excessively long newsletter and claiming there's relevant substance in there somewhere? Or at least tell me if I've incorrectly interpreted the "Meshtastic Is Rediscovering Lessons (Already Learned) of Amateur Radio Data Networking" section as listing problems but no solutions aside from buying a radically different (more expensive and power-hungry) type of radio? Try making some specific suggestions for what Meshtastic is doing wrong that could be done differently. That way, we can tell whether your beef is with the Meshtastic software and protocol, or with their choice of LoRa radio hardware, or if you're just trying to preach about your ideal mesh network design with unstated assumptions about the priorities and constraints of such a network.
- immibis 9mo ago
  All the information you seek is found in the article you stubbornly refuse to read.
  - Groxx 9mo ago
    having read that meshtastic section: I mostly agree with their requests tbh. the only suggestions in there seem to be "use full duplex" (with approximately one reason why, though it's a good one) and "solve frequency discovery with SDR" which they've already addressed as somewhat ridiculous - because it is, for someone interested in a low power and low cost network. particularly the SDR stuff, which is the VAST majority of that section. this is not at all the same target audience as meshtastic: >A computer with “sufficient” compute power and RAM, to run the ka9q-radio software. KA9Q has stated that a Raspberry Pi 4 is sufficient, and now we have the Raspberry Pi 5 with up to 16 GB of RAM, for only $120. that's like suggesting the way to fix a wireless problem is to use a wire. otherwise the criticism seems to summarize as "it's slow and bad" and well. okay? that's hardly constructive, whether or not it's accurate. the whole thing reads like "the solution is left as an exercise to the reader ;)" because it sounds like it's written by and for people who are already experts and just want to read a cathartic list of flaws they already know. and/or "buy better hardware lol". it's not at all the logical slam-dunk that you seem to think it is.
    - bb88 9mo ago
      Meshtastic can only use one frequency at a time. So, say, a battery status update can stomp on a message trying to get to a meshtastic router. (He's got the link to the hidden node problem with a great wikipedia article about it). The more popular the network, the more frequent these message stomps happen. Flood routing makes these stomps more frequent. There is also no end to end packet acknowledgement system like tcp, so at hop 3 (e.g.) if the message got stomped on, who would know? Let's say someone made a dual band lora transceiver. Well that would help, but it wouldn't solve anything else, because there's still core routing/reliability/topology issues. So if you had 20 channels to talk over, well that would be even better. The chance of having your message stomped on would go down significantly making the network much more reliable. That's the SDR part (the listening of 20 channels at once) vs the Lora chip which can only listen/transmit on 1 channel at once. Edited to add: "But that's super expensive hardware/engineering to do that!" you might say. Well, it's being done today. The point is that if you can fit 20 1khz channels in a 20khz RF space. The 20khz RF can be converted into audio and fed into a soundcard and processed. This exists today with FT-8, though FT-8 uses 150hz bandwidth per stream in 2.8khz sections per band. You can see some FT-8 activity by looking at some websdrs. Maybe go here and tune to 14.074Mhz Upper Side Band (USB) http://data3.caprockweather.com:8073/ http://data3.caprockweather.com:8073/ Each vertical line is one message 150hz wide.
      - Groxx 9mo ago
        yeah, that's exactly what I was referring to as "use full duplex" (use at least 2 frequencies, I agree this sounds like pretty solid critique (particularly with meshcore's network setup) and wouldn't make hardware dramatically more expensive) and "buy better hardware lol" (use 20+ frequencies and make it a completely different product at a MASSIVE price increase. why not just suggest wires then). so there's one bit of probably-usable advice (slightly raise cost for significant benefits) and one that completely misses the point (charge at least 5-10x more for wildly different hardware, use hundreds of times more power, etc). the article spends several times more text on the latter. flood routing and lack of end to end ack I also agree with, I sincerely doubt those are the best options and if user complaints are any sign then I think it's an existence proof that it doesn't scale, exactly as predicted. neither are part of that article, though it is in a linked also-large mastodon thread, which has basically just one constructive suggestion (8x the channels, though they don't think it'll work either) and many "this sucks" examples (flood fill, hop limit, etc. it amounts to "do better", not "X is better, learn from it").
- hedgehog 9mo ago
  FWIW I have some background in this area and got curious how Meshtastic works so I read some of the docs and code. It seems like they are unaware of existing work even 20+ years ago, a specific suggestion is to study the state of single radio CSMA meshes in say 2005 and make a list of subjects to read on, then do that. There's a lot of stuff that happened later but in the early 2000s many people tried to make meshes out of 802.11b IBSS and a lot got written about those efforts.
mtlynch 9mo ago
I've been spending a lot of time experimenting with and learning about Meshtastic and MeshCore recently,[0] and I'm also puzzled by the criticism of Meshtastic. In the article you linked, there are three paragraphs about Meshtastic in a 150-paragraph newsletter about several topics. The criticism seems to be that they they use digipeating, and then it refers to a Fedi thread[1] which is more coherent but still fairly vague. The upshot seems to be that flood routing doesn't scale, which is a fair criticism but feels disproportionate to the level of vitriol against the project. The Fedi thread also adds that the Meshtastic founders were rude or unprofessional to him but doesn't cite any specifics or evidence. I see this a lot with Meshtastic. People keep saying the founders are toxic and disrespectful of the community but it's always in these vague terms so I don't know what's driving it. But specifically in this thread, I agree with sibling poster that you're being disrespectful and arguing ineffectively by pointing to such poor resources and then blaming other people for being unconvinced or confused. [0] https://mtlynch.io/first-impressions-of-meshcore/ https://mtlynch.io/first-impressions-of-meshcore/ [1] https://partyon.xyz/@nullagent/113861754522594610 https://partyon.xyz/@nullagent/113861754522594610
- threemux 9mo ago
  As I understand it, the section on "what not to do" features many things that Meshtastic does, though it does not say that explicitly. Perhaps the linked post wasn't clear to non hams (it is a newsletter targeted at hams), but the biggest issue is not flood routing, but using the same channel for networking and user access. It, by definition, cannot scale meaningfully. Many commercial networks solve this with either FDMA or TDMA. Elsewhere in the newsletter, the author advocates for a form of FDMA, where users operate on different, dynamically allocated frequencies and all of them are received at once. P25 trunked radios used by almost all law enforcement in the US operate on a system like this. I think the vitriol from those who are in the space either professionally or as an amateur comes from the fact Meshtastic is repeating mistakes we knew about in the 80s at the latest, for which reams if literature freely exists.
  - mtlynch 9mo ago
    Thanks! I appreciate your more accessible explanation.

by bb88 9mo ago

wtallis 9mo ago

How about you make an actual argument here in this thread, instead of vaguely gesturing at an excessively long newsletter and claiming there's relevant substance in there somewhere? Or at least tell me if I've incorrectly interpreted the "Meshtastic Is Rediscovering Lessons (Already Learned) of Amateur Radio Data Networking" section as listing problems but no solutions aside from buying a radically different (more expensive and power-hungry) type of radio? Try making some specific suggestions for what Meshtastic is doing wrong that could be done differently. That way, we can tell whether your beef is with the Meshtastic software and protocol, or with their choice of LoRa radio hardware, or if you're just trying to preach about your ideal mesh network design with unstated assumptions about the priorities and constraints of such a network.

immibis 9mo ago
All the information you seek is found in the article you stubbornly refuse to read.
- Groxx 9mo ago
  having read that meshtastic section: I mostly agree with their requests tbh. the only suggestions in there seem to be "use full duplex" (with approximately one reason why, though it's a good one) and "solve frequency discovery with SDR" which they've already addressed as somewhat ridiculous - because it is, for someone interested in a low power and low cost network. particularly the SDR stuff, which is the VAST majority of that section. this is not at all the same target audience as meshtastic: >A computer with “sufficient” compute power and RAM, to run the ka9q-radio software. KA9Q has stated that a Raspberry Pi 4 is sufficient, and now we have the Raspberry Pi 5 with up to 16 GB of RAM, for only $120. that's like suggesting the way to fix a wireless problem is to use a wire. otherwise the criticism seems to summarize as "it's slow and bad" and well. okay? that's hardly constructive, whether or not it's accurate. the whole thing reads like "the solution is left as an exercise to the reader ;)" because it sounds like it's written by and for people who are already experts and just want to read a cathartic list of flaws they already know. and/or "buy better hardware lol". it's not at all the logical slam-dunk that you seem to think it is.
  - bb88 9mo ago
    Meshtastic can only use one frequency at a time. So, say, a battery status update can stomp on a message trying to get to a meshtastic router. (He's got the link to the hidden node problem with a great wikipedia article about it). The more popular the network, the more frequent these message stomps happen. Flood routing makes these stomps more frequent. There is also no end to end packet acknowledgement system like tcp, so at hop 3 (e.g.) if the message got stomped on, who would know? Let's say someone made a dual band lora transceiver. Well that would help, but it wouldn't solve anything else, because there's still core routing/reliability/topology issues. So if you had 20 channels to talk over, well that would be even better. The chance of having your message stomped on would go down significantly making the network much more reliable. That's the SDR part (the listening of 20 channels at once) vs the Lora chip which can only listen/transmit on 1 channel at once. Edited to add: "But that's super expensive hardware/engineering to do that!" you might say. Well, it's being done today. The point is that if you can fit 20 1khz channels in a 20khz RF space. The 20khz RF can be converted into audio and fed into a soundcard and processed. This exists today with FT-8, though FT-8 uses 150hz bandwidth per stream in 2.8khz sections per band. You can see some FT-8 activity by looking at some websdrs. Maybe go here and tune to 14.074Mhz Upper Side Band (USB) http://data3.caprockweather.com:8073/ http://data3.caprockweather.com:8073/ Each vertical line is one message 150hz wide.
    - Groxx 9mo ago
      yeah, that's exactly what I was referring to as "use full duplex" (use at least 2 frequencies, I agree this sounds like pretty solid critique (particularly with meshcore's network setup) and wouldn't make hardware dramatically more expensive) and "buy better hardware lol" (use 20+ frequencies and make it a completely different product at a MASSIVE price increase. why not just suggest wires then). so there's one bit of probably-usable advice (slightly raise cost for significant benefits) and one that completely misses the point (charge at least 5-10x more for wildly different hardware, use hundreds of times more power, etc). the article spends several times more text on the latter. flood routing and lack of end to end ack I also agree with, I sincerely doubt those are the best options and if user complaints are any sign then I think it's an existence proof that it doesn't scale, exactly as predicted. neither are part of that article, though it is in a linked also-large mastodon thread, which has basically just one constructive suggestion (8x the channels, though they don't think it'll work either) and many "this sucks" examples (flood fill, hop limit, etc. it amounts to "do better", not "X is better, learn from it").
hedgehog 9mo ago
FWIW I have some background in this area and got curious how Meshtastic works so I read some of the docs and code. It seems like they are unaware of existing work even 20+ years ago, a specific suggestion is to study the state of single radio CSMA meshes in say 2005 and make a list of subjects to read on, then do that. There's a lot of stuff that happened later but in the early 2000s many people tried to make meshes out of 802.11b IBSS and a lot got written about those efforts.

mtlynch 9mo ago

I've been spending a lot of time experimenting with and learning about Meshtastic and MeshCore recently,[0] and I'm also puzzled by the criticism of Meshtastic. In the article you linked, there are three paragraphs about Meshtastic in a 150-paragraph newsletter about several topics. The criticism seems to be that they they use digipeating, and then it refers to a Fedi thread[1] which is more coherent but still fairly vague. The upshot seems to be that flood routing doesn't scale, which is a fair criticism but feels disproportionate to the level of vitriol against the project. The Fedi thread also adds that the Meshtastic founders were rude or unprofessional to him but doesn't cite any specifics or evidence. I see this a lot with Meshtastic. People keep saying the founders are toxic and disrespectful of the community but it's always in these vague terms so I don't know what's driving it. But specifically in this thread, I agree with sibling poster that you're being disrespectful and arguing ineffectively by pointing to such poor resources and then blaming other people for being unconvinced or confused. [0] https://mtlynch.io/first-impressions-of-meshcore/ https://mtlynch.io/first-impressions-of-meshcore/ [1] https://partyon.xyz/@nullagent/113861754522594610 https://partyon.xyz/@nullagent/113861754522594610

threemux 9mo ago
As I understand it, the section on "what not to do" features many things that Meshtastic does, though it does not say that explicitly. Perhaps the linked post wasn't clear to non hams (it is a newsletter targeted at hams), but the biggest issue is not flood routing, but using the same channel for networking and user access. It, by definition, cannot scale meaningfully. Many commercial networks solve this with either FDMA or TDMA. Elsewhere in the newsletter, the author advocates for a form of FDMA, where users operate on different, dynamically allocated frequencies and all of them are received at once. P25 trunked radios used by almost all law enforcement in the US operate on a system like this. I think the vitriol from those who are in the space either professionally or as an amateur comes from the fact Meshtastic is repeating mistakes we knew about in the 80s at the latest, for which reams if literature freely exists.
- mtlynch 9mo ago
  Thanks! I appreciate your more accessible explanation.