4 ms·
Yarvin was quite explicit about creatively embedding neo-reaction in the design. The most obvious is the hierarchical finite namespace -- you receive your iden
by michael1999 6y ago
Yarvin was quite explicit about creatively embedding neo-reaction in the design. The most obvious is the hierarchical finite namespace -- you receive your identity from an authority, and the authorities (read as "land") are already assigned by the founders (read as landlords). The early writings were very explicit with language like "planets", "moons", etc. https://urbit.org/blog/the-urbit-address-space/ https://urbit.org/blog/the-urbit-address-space/
- Garlef 6y agoAre there ETH transaction fees of a node and do they trickle up in the id hirarchy?
- joshuakelly 6y agoThe address space is scarce (not dissimilarly to ipv4). Ethereum is used to map address space to owners and activities like transferring ownership. Urbit processes listen to an Ethereum gateway to learn about ownership changes and address lifecycle events ("breaches"), there aren't other "transactions"/fees involved in the protocol/implementation. This part of the project is called Azimuth - and can be used for other things: https://github.com/urbit/azimuth https://github.com/urbit/azimuth
- x86ARMsRace 6y agoDon't forget that in the initial source-code, Yarvin called the planets and moons, dukes and earls. To me, that's a pretty clear indicator that feudalism was on his mind at the time of developing the software.
- d_theorist 6y agoBy this argument, it is also impossible to separate DNS from Yarvin’s ideology, since that has a practically identical hierarchical system.
- api 6y agoI find the network and addressing model to be the worst designed and least interesting part of Urbit. That's the part that reflects Yarvin's political ideology, but I also think neoreaction is political flat Eartherism. Pretty much all of neoreaction boils down to refusing to account for survivorship bias. The average old thing looks better than the average new thing because crappy old things were forgotten. The interesting parts are the functional programming approach used, which is somewhat novel or at least novel in its execution. Another reply compared it to DNS, which is valid, but DNS does not have artificial limitations like Urbit does. DNS is also in many ways a bad design, not because it's inefficient but precisely because it encourages a feudal model that artificially constrains what the Internet could do. DNS was maybe the only viable model back when the largest machines online came with a whopping few megabytes of RAM, but some kind of distributed replication approach could be done today.