Oddbean new post about | logout
 Before the 1990s most computing was done on mainframes that were expensive and difficult to access, because that was simply the state of hardware.

Then we had a couple decades of software running on personal computing hardware because networking hardware sucked. 

Now with broadband we're back to running software on servers because cloud computing has made it cost efficient, and software companies can have far greater control over the user experience. 

The looming question in my mind: can we swing the pendulum back in the other direction by focusing on development of p2p apps and infrastructure? 
 The Session folks have penned the name 'Decentralized Physical Infrastructure Network', but perhaps we could work on the title.

Regardless, having Bitcoin nodes and other services, such as invidious, Cryptpad, and many others (like what  @Seth For Privacy runs) by a greater number of us would help to push the pendulum in the other direction.

The Start9 folks are doing a great job, but what else would be required to get the average user up and running? 
 I think one problem is just usability, decentralized solutions tend to either suck or need tons of maintenance, possibly both. 

We need to bring the UX and ease of use closer to entrenched centralized solutions.

Also i dont necessarily think people need to run their own stuff in a homelab at home. What we need is democratised compute - push the boundaries of confidential computing, FHE etc and create solutions where people can run encrypted blobs anywhere and everywhere without giving up the privacy and security. This way you could keep the convenience and gain decentralization 
 Let's create a hello world and go from there. 
 I can put your pub key in my ssh authorized_keys file... Send the"hello" thread to me, the "world" thread to another. 
 I wonder if the hardware is the driving force here.  Switch back to cloud computing could have also happened because of the success of the ad based profit model, and the 'free' services that came with that.  
 Need to focus on making user set up as one click as possible. Cloud is over engineered and overpriced for 90% of applications. People only use it because "it's easy" 
 Yet to most people, "being easy" is the first and most important requirement, when trying out new software/services.

As much as one might not like it, in order to onboard normies, ease of use is inevitable... 🤷🏻‍♂️ 
 That was my point... Make self hosting as easy as using a cloud service.  
 💯 
 A tangent to what you are talking about…
All these kids attending public schools are trapped in the Google ecosystem from the moment they start kindergarten. I don’t think administrators, school boards, and parents understand the implications.  I wonder if P2P can succeed at scale without some kind of awakening. 
 We should own our data and possession is 9/10 of the law. I’m in favor of this. 
 Efficiency of computing and decentralization are tradeoffs. Decentralisation is not efficient , never will be, it is secure and resilient to resist central control. This is why bitcoin will never have the throughput of centralised computing, all the nodes globally, ideally, must see and verify each transaction and there are finite limits to achieving this. When people try and disolve these physical limitations we end up with a shitcoin. In 15 years 100s of thousands of shitcoins have tried and failed to recobcile this yet we still waste time moping  about it. Centralise small payments , decentralise the base money and the big payments. Trade convenience for security, that the bottom line. Use both in its respective place. 
 Yes, @Start9 is doing amazing things with their private servers allowing anybody, no matter if you don't have a technical background (like myself) to greatly reduce your dependancy on BigTech 
 I believe https://holepunch.to/ fixes this. 

https://keet.io/ is the first app on it. A rapidly improving demo. 

We're early. In 5 years I see most client/server and SaaS apps running on holepunch. 
 That “couple decades of software running on personal computing hardware” was the time when whole families had only one computer in their household. Maybe two if you’re lucky. And certainly no smartphone.

Now, every person has their own laptop, phone, probably tablet, and maybe watch. And they expect them all to access the same data, whether they’re at home or on the go. 

Networking might be better now, but proper network security is far too complex for your average (or even above-average) person to host a home server that can be safely accessed externally. 

And the software isn’t really there yet either. Projects like #start9 help, but the big corporate services (Google, Amazon, Microsoft, Facebook, etc.) won’t give up easily and will always have better funding than any open source project. 

Doesn’t mean we shouldn’t try. I think it’s a worthy goal.

Tl;dr: Yes, the hardware is good enough. No, the ecosystem is not ready for it. 
 I just miss running applications on my computer instead of running a browser that runs a framework that runs the front end of an application that perceptibly lags because it has to consult a backend run elsewhere by a company that requires an account and sells your data. 
 I fondly remember Encarta, which was a locally run Wikipedia. Having the ability to get information on mostly anything before the internet took over felt like a superpower. 
 please get a lightning address so I can zap this  
 Mesh networks are a nice step in the right direction. There is definitely still a lot of work to do on making it as easy as possible for anyone to self-host, though. 
 You're the ETHhead. You tell me.  
 Start9 feels like a good first step in this direction. There’s really cool stuff I couldn’t do without it 
 I hope so but the challenge imo is that average people don’t care about privacy censorship or unfair business practices until it personally affects them. I’m not saying they can’t be persuaded but it’s like telling someone that likes cheap cigarettes they’re bad for you 
 Mainframes remain a backbone for financial tx processing today (>85%) because they’re cost effective for very high base load

For loads where there’s significant variance, “cloud” has gained significant foothold because it can be (not is) more cost effective

The important difference is that cloud is just someone else’s data centre over which you have zero control, AND one which incentivises centralising control and access to those hosted instances, machines and applications

Running your own servers reliably effectively and efficiently is becoming easier with tools like Proxmox .. but infrastructure remains a pita

Highly analogous to what we’re observing with ⚡️ and LSPs and the tradeoffs for efficiency, ease of use, and risk of capture 
 We need create or make more ways to communicate data via other infra that not internet only. Mash, Satellite, Radio, "Internet Nodes" (using the combination of all the 3 first ones maybe with a low cost tech that can became viral and easy to buy everywhere).

We need more Linux/decentralized OS's.

We need more IPFS solutions (decentralized media/data), more Ordinals solutions (optmized onchain data), using L1 and L2 structures... mixing it more easily and private. Connecting all it with more P2P tools. 

We need marketplaces and hosting server with free protocol like Nostr and a decentralized or immutable data storage like IPFS/Ordinals.

We are in the way but is still incomplete, IMO. 
 Everything is a choice.
Which means self discipline and collective action = change.