Item — HN Reader

I'm optimistic about self-hosting/self sovereignty (which both fall under the umbrella of what I call indie hosting) long term.

But I think both of these articles gloss over the fact that end-to-end encryption has never been shown to work in a real system with normal people. Key management is a completely unsolved problem.

If you don't have e2ee, with current tooling most people will need someone they trust to run their server. But then you run into a privacy paradox: most people have more content they would rather have google looking at/training on than someone close to them looking at, than the other way around.

Personally I think the next step forward is improving software to be more turnkey so everyone can run their own as a GUI app on an old laptop or phone.

That said, we definitely need protocols for sharing stuff.

4 comments

> end-to-end encryption has never been shown to work in a real system with normal people

I would argue that Signal is a great example of this working quite well, and tons of normal people use Signal. It's no more frictional than WhatsApp.

Signal is a promising example, but ephemeral messaging is a very different use case from things like photo management.

Signal has a lot of problems with changing devices while preserving history. As in it's often just impossible.

Are you aware of this https://signal.org/blog/a-synchronized-start-for-linked-devi...

I think so, this looks like the feature where adding an new linked device transfers messages from your phone.

What still isn't possible as far as I am aware is transferring messages from Android to iOS or vice versa. Last time I looked into this was a few weeks ago.

These are all solved problems depending on what someone is after.

Tools like tailscale/headscale combined with proxmox give most people point and click self hosting close to using a digital ocean droplet (which should never be used in production).

Tailscale + Proxmox is still an order of magnitude more difficult than this is going to need to be.

One can follow a video in about an hour maybe two and be up and have the basics down with proxmox.

Of course, if someone wants to figure out perfection the first time it might take a bit longer.

Like I said, an order of magnitude harder than it should be.

It should be install app, do an oauth flow to open a tunnel, done.

Ah, for Tailscale inside Proxmox I'm assuming.

This script exactly works like as you're describing:

https://community-scripts.github.io/ProxmoxVE/scripts?id=add...

Most humans don't know what a script is

What should never be used in prod?

Digital Ocean. Notes in my previous comments somewhere, I can see if I can find it.

Among other things, their system has a massive bug that will delete your servers without notifying anyone with a valid and working CC on file.

If Self-hosting doesn't have dissident-level privacy it's still dramatically better than centralized SaaS.

> end-to-end encryption has never been shown to work in a real system with normal people

Matrix has solved this problem.

I've tried to adopt Matrix multiple times over years.

IMO the UX is evidence that e2ee is not solved.

we solved it in element x, but there appears to be massive inertia towards adopting it because it doesn’t have threads or spaces yet.

If I only use 1 device for 5 years, then lose it, how do I recover?

From a backup?

Backed up from what? People only use the cloud for backups, at which point the e2ee is compromised.

It doesn't matter what people are currently doing. In Matrix, there is a possibility to make reliable, encrypted backups. It can also be done with a simple interface.