Centralization, Decentralization, and Internet Standards(rfc-editor.org)
rfc-editor.org
Centralization, Decentralization, and Internet Standards
https://www.rfc-editor.org/rfc/rfc9518.html
23 comments
Agreed. Developers - and by extension communities of them - tend to come up with technical solutions to problems, regardless of whether the problem is technical in nature, which invariably either don’t work at all, or make the situation actively worse when the technical solution is weaponised by those at fault in the first place.
I’m not sure how we address this, which is a pity because even an uncertain response would be better than the “confident technical solution” we often get thrown around.
I’m not sure how we address this, which is a pity because even an uncertain response would be better than the “confident technical solution” we often get thrown around.
"how" people trying their hand at "why" problems.
Are there solutions to this problem?
You’re trying to constrain meaning to your sensibilities. An authority you don’t have.
Note humanity does not use Latin in day to day life.
Language is not immutable physics.
Note humanity does not use Latin in day to day life.
Language is not immutable physics.
I disagree, it's in a large part a computer science problem.
So for instance, I work on a decentralized VR world (https://overte.org/). Our architecture is very similar to that of the web -- we have a 3D analog of a web server, and a 3D analog of a web browser. Authentication is like on the web -- anonymous by default. Storage is wherever you want to fetch stuff from by HTTP.
That's technically different from other systems in the same space. Eg, Second Life is a very centralized system. There's a grid, the various simulators are placed on a map, there's a central asset storage, and everyone must log in.
If Second Life open sourced their backend, then the resulting outcome would be competing centralized platforms, where you either log into SL, or you log into the alternative. Each of those is a self-contained universe that you can be completely be removed from.
And sure, there are various political implications there, but it's also a matter of what technical decisions and tradeoffs you make. SL for instance assumes everyone has an user account. Our system assumes everyone is anonymous by default. In SL there can be a coherent notion of a "world map" because a central entity decides which simulator goes where. In our system, such a thing is not possible because any given world has an arbitrary size and arbitrary owner, there's no way to say Alice and Bob are neighbors, nor there is a central system to ask for a definitive answer.
The system is so well designed for this that IMO it likely was one of the things that made the business model of the original maker of it (High Fidelity) fail. The design is just naturally resistant to monetization. To the point that when I first encountered it, I was baffled at how it was ever going to make any money. The very design looks like something a highly ideological and anti-corporate group would come up with to make sure that even if some of them fell to the "dark side", there still wouldn't be a way to make a profit.
So for instance, I work on a decentralized VR world (https://overte.org/). Our architecture is very similar to that of the web -- we have a 3D analog of a web server, and a 3D analog of a web browser. Authentication is like on the web -- anonymous by default. Storage is wherever you want to fetch stuff from by HTTP.
That's technically different from other systems in the same space. Eg, Second Life is a very centralized system. There's a grid, the various simulators are placed on a map, there's a central asset storage, and everyone must log in.
If Second Life open sourced their backend, then the resulting outcome would be competing centralized platforms, where you either log into SL, or you log into the alternative. Each of those is a self-contained universe that you can be completely be removed from.
And sure, there are various political implications there, but it's also a matter of what technical decisions and tradeoffs you make. SL for instance assumes everyone has an user account. Our system assumes everyone is anonymous by default. In SL there can be a coherent notion of a "world map" because a central entity decides which simulator goes where. In our system, such a thing is not possible because any given world has an arbitrary size and arbitrary owner, there's no way to say Alice and Bob are neighbors, nor there is a central system to ask for a definitive answer.
The system is so well designed for this that IMO it likely was one of the things that made the business model of the original maker of it (High Fidelity) fail. The design is just naturally resistant to monetization. To the point that when I first encountered it, I was baffled at how it was ever going to make any money. The very design looks like something a highly ideological and anti-corporate group would come up with to make sure that even if some of them fell to the "dark side", there still wouldn't be a way to make a profit.
I am sympathetic to your structural take, and also to the parent comment
that technologists lack political chops. But I lost you here:
> something a highly ideological and anti-corporate group would come up with to make sure that even if some of them fell to the "dark side", there still wouldn't be a way to make a profit.
No. The motive is to make sure that even if some of them fell to the "dark side", there wouldn't be a way to destroy the efforts of those who choose not to make a profit.
> something a highly ideological and anti-corporate group would come up with to make sure that even if some of them fell to the "dark side", there still wouldn't be a way to make a profit.
No. The motive is to make sure that even if some of them fell to the "dark side", there wouldn't be a way to destroy the efforts of those who choose not to make a profit.
That part is not really serious. The point is that this was a commercial company, of a decent size and investment. And somehow it produced a design that one would expect to come out from some sort of highly anti-corporate group.
Normally a company has a business plan of some kind. This design is almost as if somebody put a lot of thought into making sure there couldn't be one.
Normally a company has a business plan of some kind. This design is almost as if somebody put a lot of thought into making sure there couldn't be one.
I see your point that as an established commercial company they
apparently failed to diligently research their returns.
> somebody put a lot of thought into making sure there couldn't be one
Perhaps nobody put thought into that. Maybe they're just bad business thinkers. But equating openness to an ideological anti-corporate bent seems a little strong. Is it possible they saw a profitable path that is also open?
> somebody put a lot of thought into making sure there couldn't be one
Perhaps nobody put thought into that. Maybe they're just bad business thinkers. But equating openness to an ideological anti-corporate bent seems a little strong. Is it possible they saw a profitable path that is also open?
Well, take Second Life, VR Chat, NeosVR, Resonite, etc. They all are pretty similar concept-wise. First thing you get is a login screen. You need an account. On some, that account costs (or used to) cost money.
The world is locked inside a walled garden. If you don't like SL's rules, you don't get to touch anything SL related. You can't talk to your SL friends without logging into SL.
The server code is closed source. You must use the central infrastructure. In some cases, like SL, the client is open.
They sell add-ons like storage space for your stuff, and things like avatars you can directly buy. Some have an internal economy, but eg, Second Life quickly moved to being the official owners of the internal marketplace, and tax transactions.
If you want to have a permanent chunk of the world in SL, you pay for that, monthly.
All pretty standard. Now compare High Fidelity.
You install the client, open it, and instantly you show up in-world with no login screen required.
For data storage the solution is to pay any random web provider, or use the free AWS S3 tier. Any asset is just an URL pointing to https://example.com/file.gltf -- served from any random server, downloaded client-side, and outside HiFi's control.
Okay, so you pay for hosting your own environment? Nope. You host it on any Windows or Linux host you like, or the installer will even automatically install a server locally, easily start it from a menu, and NAT hole punch so you can host stuff on your laptop if you want to.
Tax Alice selling stuff to Bob? Well, the system is so open that anyone can set up their own shop if they want to. The source code for the shop is open, so is both the server and client. Since assets are hosted externally, nothing prevents Alice paying Bob with Paypal, emailing a file, and escaping any taxation.
Excluding people who break the rules? No such thing, there's no "grid". You can only be banned from individual worlds, like HN blocking you off doesn't mean anything for Ars Technica.
And all of that was Apache 2.0 licensed, so anybody who wanted could absolutely set up their own commercial operation and not share any of their own improvements.
If it was me running it, I'd have put the AGPL on it to discourage competitors, and either focus completely on custom setups for whoever needs one, or try to run a very light operation subsisting from Patreon.
The world is locked inside a walled garden. If you don't like SL's rules, you don't get to touch anything SL related. You can't talk to your SL friends without logging into SL.
The server code is closed source. You must use the central infrastructure. In some cases, like SL, the client is open.
They sell add-ons like storage space for your stuff, and things like avatars you can directly buy. Some have an internal economy, but eg, Second Life quickly moved to being the official owners of the internal marketplace, and tax transactions.
If you want to have a permanent chunk of the world in SL, you pay for that, monthly.
All pretty standard. Now compare High Fidelity.
You install the client, open it, and instantly you show up in-world with no login screen required.
For data storage the solution is to pay any random web provider, or use the free AWS S3 tier. Any asset is just an URL pointing to https://example.com/file.gltf -- served from any random server, downloaded client-side, and outside HiFi's control.
Okay, so you pay for hosting your own environment? Nope. You host it on any Windows or Linux host you like, or the installer will even automatically install a server locally, easily start it from a menu, and NAT hole punch so you can host stuff on your laptop if you want to.
Tax Alice selling stuff to Bob? Well, the system is so open that anyone can set up their own shop if they want to. The source code for the shop is open, so is both the server and client. Since assets are hosted externally, nothing prevents Alice paying Bob with Paypal, emailing a file, and escaping any taxation.
Excluding people who break the rules? No such thing, there's no "grid". You can only be banned from individual worlds, like HN blocking you off doesn't mean anything for Ars Technica.
And all of that was Apache 2.0 licensed, so anybody who wanted could absolutely set up their own commercial operation and not share any of their own improvements.
If it was me running it, I'd have put the AGPL on it to discourage competitors, and either focus completely on custom setups for whoever needs one, or try to run a very light operation subsisting from Patreon.
I would argue VR is not a neutral example.
In the physical world, space is a limited resource that sells at a premium. You have benefit from things being in a reasonable distance for walking/traveling. Being near water has real-world benefits for transportation and trade as well as agriculture.
A virtual world is still based on fundamental spatial concepts like needing to travel, having benefit of proximity/association, and so on.
So there needs to be a specific set of decisions made on how to deal with spatial assumptions when they go counter to your design goals:
- Games have instanced areas because otherwise there would be physically too many people to move, or allow you to walk through other people to get around.
- A chat service might represent _your_ rooms off a hub, rather than a megaplex conference center of all rooms.
- Individual zones might be represented as distinct 'worlds' with portal entrances, to allow for each to have its own spatial rules (and even physics), where each has its own sense of layout/locality.
each of these has their own sets of trade-offs, but importantly you want consistency for users. That set of trade-offs literally lead to representing the world as a walled garden.
That is before you get to business interests being first and foremost around succeeding with their own products, not interoperability with nonexistent or noncooperative competitors. The market has to hit a point where vendors agree that interoperability is important enough to accept the tradeoffs that come with it, before any technical effort there can succeed.
In the physical world, space is a limited resource that sells at a premium. You have benefit from things being in a reasonable distance for walking/traveling. Being near water has real-world benefits for transportation and trade as well as agriculture.
A virtual world is still based on fundamental spatial concepts like needing to travel, having benefit of proximity/association, and so on.
So there needs to be a specific set of decisions made on how to deal with spatial assumptions when they go counter to your design goals:
- Games have instanced areas because otherwise there would be physically too many people to move, or allow you to walk through other people to get around.
- A chat service might represent _your_ rooms off a hub, rather than a megaplex conference center of all rooms.
- Individual zones might be represented as distinct 'worlds' with portal entrances, to allow for each to have its own spatial rules (and even physics), where each has its own sense of layout/locality.
each of these has their own sets of trade-offs, but importantly you want consistency for users. That set of trade-offs literally lead to representing the world as a walled garden.
That is before you get to business interests being first and foremost around succeeding with their own products, not interoperability with nonexistent or noncooperative competitors. The market has to hit a point where vendors agree that interoperability is important enough to accept the tradeoffs that come with it, before any technical effort there can succeed.
> A virtual world is still based on fundamental spatial concepts like needing to travel, having benefit of proximity/association, and so on.
Yup. Second Life heavily takes advantage of that, and sells virtual land. Which actually works, because it turns out there's such a thing as a desirable location even when it's completely artificial. For instance SL has "welcome areas" where new people show up, and depending on what you do, that may or not be a good thing.
> That is before you get to business interests being first and foremost around succeeding with their own products, not interoperability with nonexistent or noncooperative competitors. The market has to hit a point where vendors agree that interoperability is important enough to accept the tradeoffs that come with it, before any technical effort there can succeed.
IMO, interoperability in the space to a significant extent is near impossible. Yeah, on the surface we all can use JPG textures and figure out an avatar standard, but beyond that things quickly get complicated.
Second Life has its own scripting language, and its own set of limited primitive object types (eg, cube, sphere) that can be created parametrically, and nothing else is compatible with that. The different internal architectures also get expressed in the way scripting works, so it just doesn't translate.
Yup. Second Life heavily takes advantage of that, and sells virtual land. Which actually works, because it turns out there's such a thing as a desirable location even when it's completely artificial. For instance SL has "welcome areas" where new people show up, and depending on what you do, that may or not be a good thing.
> That is before you get to business interests being first and foremost around succeeding with their own products, not interoperability with nonexistent or noncooperative competitors. The market has to hit a point where vendors agree that interoperability is important enough to accept the tradeoffs that come with it, before any technical effort there can succeed.
IMO, interoperability in the space to a significant extent is near impossible. Yeah, on the surface we all can use JPG textures and figure out an avatar standard, but beyond that things quickly get complicated.
Second Life has its own scripting language, and its own set of limited primitive object types (eg, cube, sphere) that can be created parametrically, and nothing else is compatible with that. The different internal architectures also get expressed in the way scripting works, so it just doesn't translate.
I wish this conversation could continue because the question of "what
adds value in a digital space" is the big one. And, for me, ways to
add value that are ethical and empowering are a subject of great
interest. Anyway, happy and prosperous 2024 to both correspondents.
A surprising value add I find is actually a walled garden.
Many 3D/VR platforms rely on it. Eg, say you want to sell avatars. Well, an avatar is a big job which can easily cost $1K or more if it's made from scratch for you. So maybe you can make a design, have it in multiple colors, and sell for $10 a copy.
But for that to work you need to work in a system that prevents duplication, and which punishes rule breaking.
That's pretty much impossible to implement in a fully open system, and results in a weird situation where being open source and decentralized is actually seen as a downside by some users.
Many 3D/VR platforms rely on it. Eg, say you want to sell avatars. Well, an avatar is a big job which can easily cost $1K or more if it's made from scratch for you. So maybe you can make a design, have it in multiple colors, and sell for $10 a copy.
But for that to work you need to work in a system that prevents duplication, and which punishes rule breaking.
That's pretty much impossible to implement in a fully open system, and results in a weird situation where being open source and decentralized is actually seen as a downside by some users.
Has a thought today about where you're right on this. Where is the
value and why does it still not feel right for me.
Walls work both ways of course. So for the inhabitants of a walled world an element of seclusion and protection is offered if they've insecure needs. Indeed there are many network security companies providing just this model.
Where I feel friction with this is in the latitude between benevolent and malevolent protection, as in "protection racket"... to cultivate dependency.
And that there's a difference between operating "in the world", and operating in special, decoupled version of your own reality isolated from everyone else which is fragmentary. So my objection comes from there, if that makes sense? I just don't think it's good for the world if we live in bubbles, some degree of mutual exposure is healthy for society if not for business,
Walls work both ways of course. So for the inhabitants of a walled world an element of seclusion and protection is offered if they've insecure needs. Indeed there are many network security companies providing just this model.
Where I feel friction with this is in the latitude between benevolent and malevolent protection, as in "protection racket"... to cultivate dependency.
And that there's a difference between operating "in the world", and operating in special, decoupled version of your own reality isolated from everyone else which is fragmentary. So my objection comes from there, if that makes sense? I just don't think it's good for the world if we live in bubbles, some degree of mutual exposure is healthy for society if not for business,
My take would be that they set out to "minimise friction", build a
user base and monetise that later, but having never seen the
"Underpants Gnomes" episode of South Park.
One could forgive that because the "build it first, squeeze it later" model /has/ worked for other tech businesses. Twitter comes to mind.
One could build a "premium" closed ecosystem within an open ecosystem which inducts new customers in. I don't think the design was fundamentally structurally flawed, so much as they didn't follow up on the opportunities created by the initial phase.
That said, I think it's a very hard transition, even if you did plan well in advance.
One could forgive that because the "build it first, squeeze it later" model /has/ worked for other tech businesses. Twitter comes to mind.
One could build a "premium" closed ecosystem within an open ecosystem which inducts new customers in. I don't think the design was fundamentally structurally flawed, so much as they didn't follow up on the opportunities created by the initial phase.
That said, I think it's a very hard transition, even if you did plan well in advance.
> One could forgive that because the "build it first, squeeze it later" model /has/ worked for other tech businesses. Twitter comes to mind.
Twitter is a walled garden though. This is more like if Twitter built Mastodon, and nearly everyone ran their own server.
So you what, sell ads? But everyone browses from their own install which doesn't serve any.
Part of the squeezing strategy is that you have a captive audience that can be suddenly asked to pay, or watch ads and they don't want to lose their social network, so they do. But this model had no such thing.
What I heard was that their intended model was "tax the economy", but besides being trivial to work around in the architecture, every content creator I met hated the completely wide open architecture. There's no way to protect content in the system. Not only it's trivial to grab somebody else's stuff, you can't be punished for doing so.
Twitter is a walled garden though. This is more like if Twitter built Mastodon, and nearly everyone ran their own server.
So you what, sell ads? But everyone browses from their own install which doesn't serve any.
Part of the squeezing strategy is that you have a captive audience that can be suddenly asked to pay, or watch ads and they don't want to lose their social network, so they do. But this model had no such thing.
What I heard was that their intended model was "tax the economy", but besides being trivial to work around in the architecture, every content creator I met hated the completely wide open architecture. There's no way to protect content in the system. Not only it's trivial to grab somebody else's stuff, you can't be punished for doing so.
To answer from a personal point of view, I'm very, very old fashioned.
I believe that successful long term business only comes from providing
compelling value to customers, building loyalty and getting them to
willingly pay you for what you offer, not what you threaten to take
away.
Much of the "capture then exploit" thinking of the recent tech world, walled gardens and the like, is anathema to me. Once you antagonise your own customers it's a losing game. They will circumvent and subvert everything you do, and good on them for being smarter and more agile. In other words I think that whole mentality is wrong, and indeed immoral.
OTOH, that isn't to say you can't build great business in other ways, and especially within interoperable, free and open systems which also add value to the whole world as well as bring you a profitable living.
So for me, "squeezing" is about using the fertile ground you create through enabling freedom to grow something that's privately profitable. When you don't have to worry about "protecting content" and enforcing behaviour you can concentrate on creating value.
Such businesses are not super scalable, and won't make you a billionaire, or satisfy investors. But they can make you a good living and are potentially stable and sustainable. I know it doesn't fit the cultural values of Silicon Valley, but to me that's the gold standard of business.
Also my experience is that it isn't ones own customers, or peer competitors that we need protection from. It's the Microsoft, Google and Facebooks of the world who can steal your lunch with impunity, write laws, break laws... they'll steal or destroy whatever they like, so in many ways it pays to stay under their radar.
Much of the "capture then exploit" thinking of the recent tech world, walled gardens and the like, is anathema to me. Once you antagonise your own customers it's a losing game. They will circumvent and subvert everything you do, and good on them for being smarter and more agile. In other words I think that whole mentality is wrong, and indeed immoral.
OTOH, that isn't to say you can't build great business in other ways, and especially within interoperable, free and open systems which also add value to the whole world as well as bring you a profitable living.
So for me, "squeezing" is about using the fertile ground you create through enabling freedom to grow something that's privately profitable. When you don't have to worry about "protecting content" and enforcing behaviour you can concentrate on creating value.
Such businesses are not super scalable, and won't make you a billionaire, or satisfy investors. But they can make you a good living and are potentially stable and sustainable. I know it doesn't fit the cultural values of Silicon Valley, but to me that's the gold standard of business.
Also my experience is that it isn't ones own customers, or peer competitors that we need protection from. It's the Microsoft, Google and Facebooks of the world who can steal your lunch with impunity, write laws, break laws... they'll steal or destroy whatever they like, so in many ways it pays to stay under their radar.
> There are other terms the authors could use: oligopoly, oligarchy, inverted totalitarianism, plutocracy, soft despotism, rent-seeking.
Those are motivations behind centralisation, but they won't all apply to all attempts at centralisation. There may be some situations where none of those particular motivations are the apparent reason behind a particular attempt at centralisation.[0]
Using any one motivation, rather than the descriptive "centralisation", to describe the phenomena will necessarily miss instances of centralisation that the discussion should apply to. Using a long list of motivations every time would be unwieldy, and could still miss instances that should be referenced.
[0] e.g. Signal https://signal.org/blog/the-ecosystem-is-moving/
Those are motivations behind centralisation, but they won't all apply to all attempts at centralisation. There may be some situations where none of those particular motivations are the apparent reason behind a particular attempt at centralisation.[0]
Using any one motivation, rather than the descriptive "centralisation", to describe the phenomena will necessarily miss instances of centralisation that the discussion should apply to. Using a long list of motivations every time would be unwieldy, and could still miss instances that should be referenced.
[0] e.g. Signal https://signal.org/blog/the-ecosystem-is-moving/
I tend to have a sinking feeling whenever this topic comes up because there's a lot of religious fervor around federation in particular that is completely divorced from reality and practicality.
But, the section of this on federation is remarkably level-headed and pragmatic. It's not a panacea and it creates a lot of friction and new problems that otherwise wouldn't exist.
And this is before you get into the benefits for the end-user, which are typically none.
The profit motive will always build a wall--technically, operationally--or an "enclosure", if you will, legislatively or politically. The only ones committed to "openness" or "open source" or "open access" are those who have already lost in the marketplace, which is why people say things like "open source is for losers" [1].
If we look at what works it's collective ownership eg Wikipedia. As another commenter shrewdly noted, this is a political not a technical problem. It's about the workers relationship to the means of production. It's who gets to capture ("exploit") the value created. And that's not a political statement. It's just materialist analysis.
But people don't want to hear that because they see themselves as future bag-holders and collective ownership threatens their ability to one day be at the top of the ladder, even though (most likely) they'll never be at the top of the ladder.
[1]: https://siliconangle.com/2014/05/29/only-loser-vendors-are-t...
But, the section of this on federation is remarkably level-headed and pragmatic. It's not a panacea and it creates a lot of friction and new problems that otherwise wouldn't exist.
And this is before you get into the benefits for the end-user, which are typically none.
The profit motive will always build a wall--technically, operationally--or an "enclosure", if you will, legislatively or politically. The only ones committed to "openness" or "open source" or "open access" are those who have already lost in the marketplace, which is why people say things like "open source is for losers" [1].
If we look at what works it's collective ownership eg Wikipedia. As another commenter shrewdly noted, this is a political not a technical problem. It's about the workers relationship to the means of production. It's who gets to capture ("exploit") the value created. And that's not a political statement. It's just materialist analysis.
But people don't want to hear that because they see themselves as future bag-holders and collective ownership threatens their ability to one day be at the top of the ladder, even though (most likely) they'll never be at the top of the ladder.
[1]: https://siliconangle.com/2014/05/29/only-loser-vendors-are-t...
I disagree about federation having no value. If email didn't exist, we'd have to invent it.
Past discussion related to this RFC: https://news.ycombinator.com/item?id=38694004
"In this document, "centralization" is the state of affairs where a single entity or a small group of them can observe, capture, control, or extract rent from the operation or use of an Internet function exclusively."
There are other terms the authors could use: oligopoly, oligarchy, inverted totalitarianism, plutocracy, soft despotism, rent-seeking. But they didn't choose these terms, because the authors aren't political science majors, they are nerds. They're looking at the problem not as one of a larger set of economic, political and social factors, but one of a computer science problem to be solved with computer science means. If you've only ever been a carpenter, and your only tool is a hammer, every problem looks like a nail.
But this isn't a computer science problem, and none of the solutions presented in this document will prevent the forces at work from merely "routing around" any technological barriers imposed. You cannot program around a government.