A mental model for decentralization

https://jacobobryant.com/p/mental-model-decentralization/

by jacobobryant • 5 years ago

78 77 5 years ago

dkh• 5 years ago

While I agree that there should be more discussion about what should/shouldn't be decentralized and to what extent, the huge issue right now that's responsible for how heated of a debate this is right now, is in the implementation. It seems almost like it's already been decided by players in the decentralization space that the only solution is a blockchain with a token economy model that sits nicely alongside -- or in fact integrates with -- crypto trading, NFTs, etc. Now add to it the rat race everyone is in to transition to a DAO as soon possible, despite the existing insane complexity modification difficulty of blockchain platforms, or that the DAO concept has a short history but one filled with failures, and we're left with something that, frankly, frightens me.

I'm not anti-crypto, and in fact I work in the space and am primarily focused on some areas I feel are in need of being "decentralized" in the CDN and streaming video spaces. And while a ton of technical progress has been made, I can't shake the feeling that we are in many cases implementing this stuff wrong, or haven't even bothered to address the "how" or "why" in as much length as one should before going out and building it.

k__3• 5 years ago

What are good alternatives that also have useful incentivization mechanisms built in?

Every system that is lacking this is doomed.

loceng7• 5 years ago

Democratically elected government which implements policy mandating data mobility (etc) at penalty of exponentially increasing fines for non-compliance? You don't need to use a carrot for corporations, a "stick" will work just fine for making sure they follow what's determined to be ideal.

grey_earthling4• 5 years ago

Can you elaborate on what you mean by “useful incentivization mechanisms” and why they're necessary?

I use scuttlebutt because it's fun and nice. I'm not sure what more incentive is needed.

austincheney4• 5 years ago

Implementation is whatever two parties agree upon, otherwise it isn't decentralized.

grey_earthling4• 5 years ago

It seems almost like it's already been decided by players in the decentralization space that the only solution is a blockchain with a token economy model that sits nicely alongside -- or in fact integrates with -- crypto trading, NFTs, etc.

This isn't my experience. These types of solutions may be generating a lot of interest in some places (perhaps because the people talking about them hope to profit?), but they're not all that exist.

danShumway3• 5 years ago

by players in the decentralization space

Keep in mind that this is a somewhat large space, and depending on what parts of the decentralization space you're looking at you might see players who are all-in on blockchain tech, or you might see people who view it as a giant waste of time or even actively campaign against it. In particular when I look at projects that are using federation (Mastodon, Matrix, etc), I see a lot of people who are fairly skeptical about cryptocurrency and outright hostile to NFTs.

I have my own biases there, I don't think that NFTs are in practice particularly decentralized, and I've slowly come to believe that cryptocurrencies have both systemic problems that make them less valuable as a decentralization tool than they're often advertised as being, and community problems that make them difficult to use as a basis for any kind of political movement (and decentralization is inherently a political idea/movement).

But my biases aside (I'm not trying to make a pro/anti blockchain post here), just remember it's a big space without a single uniformity of views; from the circle of advocates I'm paying attention to and interacting with the most, I would have almost guessed the opposite, absent better stats I'd have guessed from my experiences that at least a plurality if not the majority of the decentralized "community" is against this stuff. Interesting to run into people who seem to be more in touch with parts of the community that are outside of my bubble.

s7r• 5 years ago

This is akin asking 'Will we build our community around hammers or screwdrivers?'

Network architectures are means, not ends.

The question is what are we trying to do?

Based on that context, we can determine whether a hammer, screwdriver, centralized architecture, decentralized architecture, or federated architecture is most beneficial for us. They all have tradeoffs (e.g. easier to remove screws than nails, easier to secure centralized than decentralized systems, easier to power federated systems than decentralized systems, easier to trust decentralized and federated systems than centralized systems) which can make them more or less useful for us in any given context. The point is, they are tools to help us realize our desired outcomes -- not the outcomes themselves.

RF_Savage• 5 years ago

Yeah. This seems to be missing from the current conversation.

0xbadcafebee2• 5 years ago

Finally, some common sense.

There's some weird psychology of tech circles that keeps us focused on decentralization. Often it just seems like bikeshedding rather than problem-solving. My guess is the Social Media generation has become wary of people controlling their lives, preferring autonomy. They hear about network decentralization, and then come to the conclusion that it must be better than the alternative, because the alternative is centralization, which is Google and Facebook (but never Apple apparently, which is pretty ironic).

austincheney4• 5 years ago

I am writing a fully decentralized application and in doing so have learned a lot about the concept. There has been a lot of interest on the subject over the last few months so I likely need to write a document about the things I have learned. Some quick things:

* Security is wildly different in a decentralized model compared with a client/server model.

* Decentralization is about one thing only: maximum autonomy.

* Federated systems are semi-autonomous. They are a move towards decentralization but not completely so. With decentralization you only need an application that send/receive agreed upon instructions.

* Anonymity and privacy are opposing qualities. Decentralization is not about either but enables one or the other.

As I have been exploring this subject I have formed some criteria for decentralization:

* Address based (IP addresses or identifiers that resolve to an IP address without third party consultation, no DNS)

* Protocol agnostic (what ever two parties can agree upon)

* No third parties (no servers or message proxies unless you own them)

Think about it in terms of snail mail. The post office moves packages around without opening or inspecting the contents. The contents are never stored in a database and could be broken. It is address to address communication.

Decentralization allows for personal or organizational devices to be remotely connected as a large physically distributed computer:

* A single file system, cross OS.

* Remote application execution, such as Steam’s remote game play.

* Content on demand. Push media out to your friends or access their computer and pull it off yourself.

loceng7• 5 years ago

"* Content on demand. Push media out to your friends or access their computer and pull it off yourself."

Reminds me of dialing into a BBS way back when, that early on was just someone with a non-dedicated or dedicated telephone line hooked up to their personal computer's modem; you could connect to fun, non-real-time multi-player text-based games (ASCII art FTW!) hosted on their PC, and some had file servers open for download and sometimes upload.

Comevius• 5 years ago

I'm writing an offline-first database and it typically uses a server for storage and coordination. The difference is that I don't think the server should be able to see what it stores and coordinates, for which searchable encryption is the solution.

However my database is address based, it reads and writes blocks of encrypted data, and it is protocol agnostic, it works over S3 and Google Drive, or could work over a pendrive sent in a letter back and forth, or IPFS. It just requires a single source of truth and optionally event notification.

So would this fit into an every peer-to-peer system is decentralized, but not every decentralized system is peer-to-peer picture, or is this not decentralized at all, because you would normally use it with Amazon S3, albeit in a maliciously-secure manner (the confidentiality, integrity and authenticity of your data is secured via authenticated encryption, the client keeps the key, the access pattern is obfuscated).

austincheney4• 5 years ago

Peer to peer systems are not necessarily decentralized. They are decentralized if their traffic does not require access by a third party. Its the difference between Napster and BitTorrent.

In the case of your scenario it could be considered decentralized as you are using cloud storage as merely a storage point, as in a dedicated redundancy, but I am hesitant to consider a dedicate server as a point of decentralization regardless of its intention or access limitations.

mlatu2• 5 years ago

considered 9p yet?

austincheney4• 5 years ago

I had no idea about this. I am looking at it now.

https://en.m.wikipedia.org/wiki/9P_(protocol)

Here is the approach I am taking: https://github.com/prettydiff/share-file-systems

k__3• 5 years ago

"how much decentralization do we need?"

I think, that hits the nail on its head.

Many people are happy with MANGA basically owning the Internet and many people aren't.

If we look at companies bending the knee for China or shutting down accounts, simply because they have different moral than the account owners, it's clear that we need much more decentralization than we have right now.

How much? I don't know.

loceng7• 5 years ago

I think part of the solution is to make sure there is a functional and tested decentralized system that's available as a fallback or failsafe, say in case the system(s) get captured by tyranny or authoritarians; and these fallback systems must be enshrined in policy/constitution, so that if they're ever attacked or targeted then that becomes a canary signal.

Data mobility could also perhaps be considered the other side of the decentralized coin.

gandutraveler• 5 years ago

Centralization is a human nature. Internet is already decentralized but we still tend to congregate in common space. Even with web3, organizations like polkadot are creating a centralized space around them. So I doubt we will ever achieve true decentralization there will always be a hierarchy and a central body that will have more control over it.

hwhelan210• 5 years ago

I don't think what we're running into here is "human nature" so much as it is investor nature. I'm not particularly familiar with them, but Polkadot has raised $300M. There has to be some return for that investment (i.e. they need to be able to lock users into their ecosystem and prove to investors that a competitor can't come along and provide an equally valuable service without starting from scratch). How can a truly decentralized service / protocol compete with the companies that are getting that kind of investment?

kuroguro• 5 years ago

there are the decentralization maximalists who think that blockchain will make tech platforms obsolete

I doubt blockchain is a silver bullet. I really liked Kleppmann's work on local-first software / CRDTs. It seems a lot closer to something practically usable at least.

https://www.youtube.com/watch?v=Qytg0Ibet2E

blamestross2• 5 years ago

Blockchain tech just offuscates centralization, it doesn't provide decentralization.

brightstep• 5 years ago

In terms used by the memo discussed by the article, blockchain are still susceptible to indirect and platform centralization.

delaaxe• 5 years ago

In other words, it’s ok for some things to be “decentralized enough”, as opposed to fully decentralized (whatever that means).

foxhop• 5 years ago

Hello, I have a similar thread going here. I think we need to "shard the centralized monolith".

https://www.unturf.com/let-us-define-a-permacomputer/

I avoid writing prescriptively to avoid setting a bias on the talent.

I AM growing my own "YouTube". Not open source yet but I'm planning big things.

https://media.unturf.com

verisimi2• 5 years ago

Here's a fantastic quote I heard this week:

"A successful utopian society would most likely be one made up of 1000s of smaller utopias."

This strikes me as the valid mega-target that we should be attempting to reach collectively - space for all.

It should also be the foundational principle for software solutions. And neutral software solutions should also be able to accommodate people who don't want to be in the proposed solution!

skohan• 5 years ago

It's a good start but you still need a mechanism for preventing negative externalities between utopias.

verisimi2• 5 years ago

The foundational principle is the same as it ever was: 'do no harm'. The golden rule.

Do not initiate harm. In reverse, do not roll over when someone is initiating harm against you.

coderzach• 5 years ago

Land Value Tax solves everything! /meme

pessimizer3• 5 years ago

Better would be 1000s of overlapping utopias. Reducing the geographic size of government seems like a lot less desirable a way to shrink it than reducing the scope of governments.

That was the original rhetoric behind "Defund" IIRC. Break up the central police force to a rump used to coordinate and regulate police functions that have been devolved to exist under other government departments. Basically a bid to put a domestically-operating military under more thoughtfully-organized civilian control instead of simply having city cops, state cops, and country cops whose only points of oversight come respectively from the mayor, the governor, and the president.

blamestross2• 5 years ago

First, thinking of decentralization/centralization as having a boundary between them can be useful, but is wrong. They are a continuum upon which things oscillate.

Decentralization fundamentally trades inefficiency for long-term robustness. Centralization trades long term robustness for short term efficiencies.

Under our implementation of capitalism, most people don't value robustness on a timescale longer than 20 years unless there was recently a crisis to remind them of it.

So every aspect of human infrastructure oscillates on the centralized-decentralized axis over time. Thinking of "decentralized" as something that will stay that way is wrong, over time people will ever centralize the system in pursuit of short term returns until the perceived "decentralized" system might as well be centralized in practice (see internet infrastructure), once a crisis happens due to over-centralization we build a new system (or re-brand an old system) that is more decentralized in response. That system eventually decays into centralization the same way.

I value changing society to think in the long term, so I think it is a lot more valuable to change the cycle versus attempt to participate in just one swing of it.

lottin3• 5 years ago

Discussions about decentralization should start by defining the term decentralization in the specific context of the discussion, otherwise nobody knows what they're talking about. For example, nobody in the economics profession would refer to the financial system as being 'centralised'. It is not. There's no central planner. Financial institutions are autonomous and make their own decisions with regards to what products to sell, and how to make them. If you want to argue that it's centralised, you will have to be much more specific. And the same applies to everything else, from the internet, money, software or whatever.

bitxbitxbitcoin2• 5 years ago

Good point. Control of the inflation rate (total money supply) in fiat currency is centralized.

pessimizer3• 5 years ago

Isn't the same thing true for bitcoin? As far as I know, it was set with the first implementation, and can only be changed when a majority of miners agree to change it.

3np• 5 years ago

Thank you. It really grinds my gears when people/projects/companies make strong claims about “decentralization” without even specifying what they’re talking about. Many times it becomes nonsensical.

pphysch• 5 years ago

For example, nobody in the economics profession would refer to the financial system as being 'centralised'. It is not. There's no central planner.

There is and it is called the Board of Governors of the Federal Reserve System. By setting the EFF rate (or QE policy if EFFR ~= 0%) of the global reserve currency, they control many facets of the financial system and broader economy.

It's not very fine grained control ("bake X loaves of bread"), but it is centralized control and planning.

Kranar2• 5 years ago

The economy as a whole isn't centralized but the commercial banking system is centralized by each nation's respective "central bank" [1]. It would be kind of awkward to claim that economists are unaware that a central bank isn't a form of centralization, specifically of a nation's monetary system.

That said, most people can pick up the meaning of centralization based on the context without every discussion having to spend a paragraph defining it. A discussion about a centralized Internet isn't too hard to understand, nor are discussions about finance, media, markets, governments, etc etc... The idea that people can't figure out what decentralization means in the context of a discussion about the Internet or finance unless someone spells it out in precise detail is kind of obtuse.

[1] https://en.wikipedia.org/wiki/Central_bank

lottin3• 5 years ago

Economists are fully aware of central banks and the role they play in the monetary and the financial system. Still they won't use the term 'centralised' because it doesn't describe the current system well. In fact, it does a terrible job of describing it. For example, consider the fact that 90-95% of the currency in circulation is not issued by the central bank, but created by commercial banks via fractional reserve banking.

vishnugupta• 5 years ago

There's no central planner

There may not be one central planner but there are definitely power centers that exert far more influence than rest of the participants combined.

There’s fed which walks a tight rope of managing inflation while keeping low unemployment rate. And then there’s federal government, by far the biggest spender in the market. You also have Fannie Mae etc selling mortgages which is huuuge.

0xbadcafebee2• 5 years ago

From the linked article Avoiding Internet Centralization:

5.2. Encrypt, Always

  When deployed at scale, encryption can be an effective technique to reduce many inherited
  centralization risks. By reducing the number of parties who have access to content of 
  communication, the ability of lower-layer protocols and intermediaries at those layers
  to interfere with or observe is precluded. Even when they can still prevent communication,
  the use of encryption makes it more difficult to discriminate the target from other
  traffic.
  
  Note that the benefits are most pronounced when the majority (if not all) traffic is
  encrypted. As a result, protocols SHOULD be encrypted by default.

They're literally describing the lack of decentralization that encryption brings. "If you encrypt, intermediaries have no control". Which means a central actor does. By encrypting communications, without an intermediary protocol to support more flexible operations, there is no way to extend functionality for networks that depend on intermediary controls.

All systems that scale HTTP-based applications require a lack of encryption on the backend. The encryption is terminated at the load balancer and the entire connection is then inspected in plaintext. Sure, the "best practice" is to encrypt between endpoints, but they must be decrypted at each hop so that load balancers, API gateways, and applications can inject or parse headers, perform rewrites, forward traffic, inspect requests (WAF), create sticky sessions, and cache media. If all HTTP connections were end to end encrypted, the web could not scale.

Another example is DNS over HTTPS. The encryption centralizes control of the protocol at the server. The encryption removes the ability of intermediate LAN/WAN DNS resolvers to cache records for a large local network, or to serve their own custom zones transparently, or do things like ad-blocking. As there are valid and necessary use cases, it means that now new "hacks" will have to be developed to accommodate the old use cases, that the protocol design intentionally ignored. The result is more centralization.

Distributed decentralized applications also suffer when PKI is used for encryption. It is very common for backend applications to stop working because they were using a private CA to encrypt backend connections and the certs expired without warning, killing all the connections. Refreshing the certs then involves re-deploying not just the servers, but all the clients that were reliant on an old client cert, and possibly an expired CA cert. This use of a private CA also breaks decentralization compared to the public CA model.

armchairhacker• 5 years ago

Honestly how decentralized is the Internet? AFAIK there are groups like W3C and ICAAN which basically control domains, and a few IPs which control routing. Most of the internet is also behind Cloudflare, AWS, etc.

The key is that these companies are very conservative when blocking users and giving others special privileges. Your service won't be blocked by Cloudflare and IPs unless it's particularly bad or your government requests it. Whereas e.g. YouTube has very strict rules and often bans people incorrectly or for debatable reasons.

Idk if the issue is actually decentralized vs. centralized. I think it's "do I decide the rules or do I just enforce them?"

golemotron• 5 years ago

Given my mental model of centralization vs. decentralization, I think it's more productive to think about the boundary: how much decentralization do we need? If there's too little, then software will stagnate from lack of competition; if there's too much, then it'll stagnate from lack of coordination. So we want to figure out where the sweet spot is.

This is a static view. A dynamic view would recognize that systems have a lifetime and tend to go through centralize and then disintegrate. Decentralization often exists prior to centralization. Catch systems on the rise.

A mental model for decentralization

Related Stories