@hazelweakly more seriously though, this connects immediately in my head with this:
https://www.theguardian.com/environment/2025/aug/02/self-termination-history-and-future-of-societal-collapse

We need to build communities you mention, ones that can support one another, to replace certain parts of the global logistics chain, so that we might… well, survive.

@fd93 There’s actually a fantastic and well studied answer for this :)

It’s about preventing “the tragedy of the commons”. Elinor Ostrom did much of her life research on answering precisely this question

https://www.youtube.com/watch?v=T6OgRki5SgM

The short answer is “there’s about 8-10 properties of a community that is long lived, stable, and communal”. Those communities naturally resist privatization, backdooring , and other abuses

Of course one can always ruin a good thing by blowing it up in an act of war… But that doesn’t mean we can’t build these communities anyway. Merely existing sustainably in of itself is resistance

@valpackett @hazelweakly it was hoped that #pgp/ #gpg would allow the development of 'Webs of Trust' which could be a key building block for this idea, but so far no one has built a user interface which would make that accessible to most people.

That UI is a hard problem to solve well, but some mashup between #ActivityPub and a web of trust -- a #Mastodon-like system where all posts are encrypted but all posts by people who trusted you are decryptable -- is possible.

https://en.wikipedia.org/wiki/Web_of_trust

@onekind Oh for sure. You often can’t actually survive without it. But most of the subsistence farming in western cultures quickly developed abusive power structures and turned into feudalism and/or slavery rather than sustainable communities.

That said: we know how to build those sustainable communities now, and we know what it looks like. All that’s left is to grow awareness and resolve around making it a reality. Of course, that’s the hard part, so I don’t want to diminish the monumental task that it will be; we’d be lucky as a species if this took less than a few centuries

@badsynthesis @hazelweakly

I would have said the opposite. End-to-end encrypted and integrity-protected storage is easy, so storing your at-rest data on someone else’s computer is pretty easy. There are even designs that are trace oblivious, so the host learns nothing from access patterns. If you want to increase reliability, RAID is old technology and splatting error correction information across multiple providers is something that we’ve known how to do for ages. Do it with enough providers and you’re resilient against one deciding to drop you as a customer.

Having a network endpoint that is reachable from the outside world, is under your control, and does not leak things like your residential address, is hard.

Running a program on a computer that other people have physical access to, with guarantees that they cannot see or tamper with it, is very hard. I think we did a pretty good job with Arm Confidential Computing Architecture, but there are a lot of corner cases that you need to be careful of even in a good implementation.

@soviut @hazelweakly

I found the article’s definition of self hosting a bit extreme. By that definition, I’ve never done it: I’ve always rented a computer or VM in someone else’s datacenter.

I can see that, as we move to FTTP being common, hosting things at home might be feasible. The dedicated server I’m renting has a 300 Mb/s network connection, I get more downstream than that and the other local FTTP provider offers more than that symmetric. But whether you’d want to is another question. Someone else pointed out that your IP for residential IPs can be fairly easily mapped to your address, and handing your home address to random Internet people is not always a great idea.

The problem, as the article identifies, is that there’s a huge gap between buying an off the shelf service and buying or renting a machine that you configure everything on. There are some somewhat appliance-like projects, but they require a lot of configuring and my experience with the ones I’ve tried is that they are worse than doing it yourself: you get a container where someone has configured things in a non-standard way and so documentation doesn’t quite work.