On 5/24/20, Karl <gmkarl@gmail.com> wrote:
A general purpose network sounds nice. Everything is doable.
What do you think of forking the codebase of an existing network, like tor or gnunet or one of the newer examples from anonymity research?
What networks ultimately do, whether they are "for" voice, video, IRC data, "messages" email files nntp etc, http style interactivity, file storage retrieval, cryptocurrency, etc, etc.... ... is move data from A to B... that's it, that's all they do. They move a "message" data, a blob of bits, a "packet" from A to B. Potentially but rarely in multicast-ish ways, sometimes in route-relay-ish ways, but always, ultimately at lowest layer, from A to B. [1] There's probably quite little return in doing all the research just to build some application specific network to be secure in just that application, because under the hood all it did was secure a more specific form of A to B for that app alone. Yet by extending the initial upfront research a bit more, you reach a general form of secure A to B for all applications, such that each new application needs to do almost no work to ride on top, applications become essentially just plugins over a data moving network, not networks themselves. Further, there is timely need metrics... plowing resources into making the best "chat" net, starves research from all other standalone app specific nets, it builds incompatible towers of networks which cannot interoperate, and they compete for exclusive node count funding etc instead of combining node count bandwidth for the commons. If need be, nodes within the commons can offer more specific transport/plugin features. Last, creating dozens of app specific nets cannot take advantage of riding and hiding in each others noise over a common transport overlay layer. And makes more risk for singled out political attack on against one app than against a general purpose net. Perhaps tor is not best as all it does is TCP. Phantom offers raw IPv6 for all existing apps, and is currently light enough that may be an ok candidate for whitepapering a dynamic chaff anti-traffic-analysis bolt on tech proof of concept, but IPv6 is not a generic data message handling network in sense of application level concepts. Building a more generic network may serve better long term, but will produce and require new apps to compile to its plugin API such as i2p-snark torrent app is specific to i2p net and has to use its API instead of just using IPv6. More generic nets could offer IPv6 as a plugin on top of them. But the extra layering to do that will make them slower than tor/phantom style IP base alone. Not to say "forking" any one net is better than other as 10 or more already extant or papered could be evaluated for that, just that a generic or at least IP API design may be more likely to produce a mass payoff effect than say building the next singularly focused impenetrable "cypherpunk mixmaster email network", which is a useless waste to anyone wanting to do browser, IRC, file, coin, voice, etc. [1] Note the only place you'll find research on anything different from A to B, is from people trying to design fundamental alternatives to IP networks... broadcast, radio, satcom etc.