3 ms·

I'm glad that more people are now thinking of mesh, or radio-delivered, internet. I think there is a lot we can do (as programmers) to start supporting mesh-ne

I'm glad that more people are now thinking of mesh, or radio-delivered, internet. I think there is a lot we can do (as programmers) to start supporting mesh-network style Internets. We're at a critical point right now where distributed systems research has turned up some decent consensus algorithms, where there's now a (finally understood) need for a computer network run by the people who use it, and where consumer radio/copper technology is fast enough to start building out reasonably fast networks.

betterunix2 9y ago
"We're at a critical point right now where distributed systems research has turned up some decent consensus algorithms," Not by a long shot. We have proof of work and systems that rely on all parties being known and identifiable, and not much else. Proof of work does not scale and the need to identify all parties requires at least some centralization. Beyond that, we also live in an age where the bulk of Internet traffic is video, and thanks to the decades-long effort of the RIAA and MPAA, most video distribution is centralized, so people demand high throughput; and the widespread use of TCP also requires low latency. Mesh networks are still not up to the task despite all the progress. You want to support mesh networks? Great! I'm with you, but let's not oversell the state of the art. Mesh networks are still experimental and without some centralized coordination (which is a de facto ISP) the performance and reliability really start to become questionable.
- gravypod 9y ago
  High throughput is 'easy' to obtain if you don't mind high latency. The only thing the internet is great at is reducing latency. Luckily for human scale actions this is not a big problem. Mesh networking technology is old at this point. Application layer software that takes advantage of it's strong suits has yet to be developed. Though there are many ideas of where to start. As for what you have brought up. In my opinion many-replica content networks, when combined with modern video compression technologies, could solve the issue of video streaming from a no-master and high-latency system.
EGreg 9y ago
Actually, I think global consensus is overused. There is no need to use consensus for most things. To achieve consensus at all costs you'd need to merge forks no matter how many updated happened in the meantime, which is usually not what you actually want. See for example https://intercoin.org https://intercoin.org project we're working on. Specifically the Technology link. Also see https://github.com/Qbix/architecture/wiki/Internet-2.0 https://github.com/Qbix/architecture/wiki/Internet-2.0 As Einstein said, you don't need 100 scientists to sign something is wrong. If they could prove it, you'd only need one. Same with gossiping a proof that some operation is not valid.

betterunix2 9y ago

"We're at a critical point right now where distributed systems research has turned up some decent consensus algorithms," Not by a long shot. We have proof of work and systems that rely on all parties being known and identifiable, and not much else. Proof of work does not scale and the need to identify all parties requires at least some centralization. Beyond that, we also live in an age where the bulk of Internet traffic is video, and thanks to the decades-long effort of the RIAA and MPAA, most video distribution is centralized, so people demand high throughput; and the widespread use of TCP also requires low latency. Mesh networks are still not up to the task despite all the progress. You want to support mesh networks? Great! I'm with you, but let's not oversell the state of the art. Mesh networks are still experimental and without some centralized coordination (which is a de facto ISP) the performance and reliability really start to become questionable.

gravypod 9y ago
High throughput is 'easy' to obtain if you don't mind high latency. The only thing the internet is great at is reducing latency. Luckily for human scale actions this is not a big problem. Mesh networking technology is old at this point. Application layer software that takes advantage of it's strong suits has yet to be developed. Though there are many ideas of where to start. As for what you have brought up. In my opinion many-replica content networks, when combined with modern video compression technologies, could solve the issue of video streaming from a no-master and high-latency system.

EGreg 9y ago

Actually, I think global consensus is overused. There is no need to use consensus for most things. To achieve consensus at all costs you'd need to merge forks no matter how many updated happened in the meantime, which is usually not what you actually want. See for example https://intercoin.org https://intercoin.org project we're working on. Specifically the Technology link. Also see https://github.com/Qbix/architecture/wiki/Internet-2.0 https://github.com/Qbix/architecture/wiki/Internet-2.0 As Einstein said, you don't need 100 scientists to sign something is wrong. If they could prove it, you'd only need one. Same with gossiping a proof that some operation is not valid.