This document offers a definition of centralization, explains why it is undesirable, identifies different types of it, catalogues limitations of common approaches to decentralization, and explores what Internet standards efforts can do to address it.
Providing definitions should be super helpful for people at my place of work as well as folks working out in the open. I wonder if and how the blockchain folks will respond to a document such as this.
Their definition of centralization is a bit muddied. For me it is a synonym of censorship, i.e. the ability to censor is a centralizing force (w.r.t. the centralization <~> decentralization continuum).
There are connections between those concepts, but they’re not synonyms. You can have a centralized system that happens not to do any censorship (e.g. something like 8chan), and you could have decentralized systems with censorship (e.g. a government forces all ISPs to make their SMTP servers block emails with certain keywords.)
it’s about the capability, not whether it is actively used (eventually it will be used to centralize, this is because I feel centralization is an economic concept and the mechanism that drives it is censorship - so the definition I like is somewhat teleological).
also; this is a bit unrelated but my experience with the chans (which is extremely limited) is that they (as a community) censor posts they don’t feel aligns with their game (because apparently if you comment on a thread it goes to the top so they just ignore things or even ask people not to comment on things .. I found out because I wanted to have a conversation with a person and use one of these public forums as a channel but yeah my experience is that the community will try their hardest to remove it)
Centralization provides a lot of capabilities. It’s also just plain easier to implement, so there’s no need to invoke economics or totalitarianism or whatever. The easiest way to communicate is client-server. The server is a singleton, so you only have to deal with keeping the source-of-truth in one place. Even with no economic interests, if you give away your code, all you get is a bunch of isolated centralized systems. Want them to interoperate? Now you have to do a bunch of much, much harder design work. And once you figure that out, you have to get all the other people to agree with your architecture and not get lost in endless bikeshedding.
I have seen this kind of stuff. It’s painful, and there doesn’t need to be any kind of profit motive or evil motivations, just human egos.
Human egos are the foundational problem of economics. Economics has nothing to do with money and everything to do with evaluation and resource allocation. Centralization is censorship not because of profit or evil but rather because of the tragedy of the commons, as soon as there is resources to be allocated (so you’ve hit capacity) then there is censorship to be deployed and yes it happens to be easy to trust someone with that job rather than share the burden (precisely due to the cognitive cost of coordination that you describe).
I do like the term precisely because it is loaded. The primary reason to care about decentralization is to avoid censorship, that is, filtering of your messages or messages you want to read by the party that controls a centralized service. Filtering messages you don’t want to receive on your own behalf is orthogonal to whether the service you use to receive messages is centralized or decentralized. In fact, centralization might make it harder to filter messages you don’t want, by imposing limits on what client software you can use to connect to a network. Anyone who complains about Discord or Twitter not giving them the tools they need to block messages they don’t want is succumbing to this problem, because Discord and Twitter are both centralized services that try to prevent users from using alternative client software.
Centralization happens because of two mutually reinforcing effects - centralized functions are more profitable to operate, and they are more convenient and cheaper to use.
I’m disappointed that their list of decentralized approaches ignores peer-to-peer systems, while including (ick) blockchains.
Providing definitions should be super helpful for people at my place of work as well as folks working out in the open. I wonder if and how the blockchain folks will respond to a document such as this.
Their definition of centralization is a bit muddied. For me it is a synonym of censorship, i.e. the ability to censor is a centralizing force (w.r.t. the centralization <~> decentralization continuum).
There are connections between those concepts, but they’re not synonyms. You can have a centralized system that happens not to do any censorship (e.g. something like 8chan), and you could have decentralized systems with censorship (e.g. a government forces all ISPs to make their SMTP servers block emails with certain keywords.)
it’s about the capability, not whether it is actively used (eventually it will be used to centralize, this is because I feel centralization is an economic concept and the mechanism that drives it is censorship - so the definition I like is somewhat teleological).
also; this is a bit unrelated but my experience with the chans (which is extremely limited) is that they (as a community) censor posts they don’t feel aligns with their game (because apparently if you comment on a thread it goes to the top so they just ignore things or even ask people not to comment on things .. I found out because I wanted to have a conversation with a person and use one of these public forums as a channel but yeah my experience is that the community will try their hardest to remove it)
Centralization provides a lot of capabilities. It’s also just plain easier to implement, so there’s no need to invoke economics or totalitarianism or whatever. The easiest way to communicate is client-server. The server is a singleton, so you only have to deal with keeping the source-of-truth in one place. Even with no economic interests, if you give away your code, all you get is a bunch of isolated centralized systems. Want them to interoperate? Now you have to do a bunch of much, much harder design work. And once you figure that out, you have to get all the other people to agree with your architecture and not get lost in endless bikeshedding.
I have seen this kind of stuff. It’s painful, and there doesn’t need to be any kind of profit motive or evil motivations, just human egos.
[Comment removed by author]
Human egos are the foundational problem of economics. Economics has nothing to do with money and everything to do with evaluation and resource allocation. Centralization is censorship not because of profit or evil but rather because of the tragedy of the commons, as soon as there is resources to be allocated (so you’ve hit capacity) then there is censorship to be deployed and yes it happens to be easy to trust someone with that job rather than share the burden (precisely due to the cognitive cost of coordination that you describe).
When you express it that way, I agree with you :)
I dislike the term “censorship” in this situation as it is very loaded. I’d prefer the term “filtering” instead.
I do like the term precisely because it is loaded. The primary reason to care about decentralization is to avoid censorship, that is, filtering of your messages or messages you want to read by the party that controls a centralized service. Filtering messages you don’t want to receive on your own behalf is orthogonal to whether the service you use to receive messages is centralized or decentralized. In fact, centralization might make it harder to filter messages you don’t want, by imposing limits on what client software you can use to connect to a network. Anyone who complains about Discord or Twitter not giving them the tools they need to block messages they don’t want is succumbing to this problem, because Discord and Twitter are both centralized services that try to prevent users from using alternative client software.
Centralization happens because of two mutually reinforcing effects - centralized functions are more profitable to operate, and they are more convenient and cheaper to use.