alt Hacker NewsI'm surprised people are advocating self-hosting as a viable solution. It takes a lot of knowledge to do sync and backup yourself, most of it implicit knowledge that people here don't realize we have and so for us it seems very easy.
There was a comment in another post on the front page about how anyone "remotely technical" can set up a docker container, and I think this is a good example because the mechanics of it are simple (edit a couple text files, run a couple commands), but half the world couldn't tell you what a terminal is and they're focused on other things in life instead of learning how computers work. Cloud succeeded because cloud is easy (at least in the beginning), it's that simple.
If we are to solve this problem, we're going to have to make self-hosting easy enough for the average 7-8 year old to do it without struggling. One promising way forward is with local-first E2EE sync and backup. The only good implementation I know of personally is Obsidian Sync, which has a UX that I adore, and hope to see more of in the future. There's other good options too, but none that I'd feel comfortable trusting a seven-year-old to execute correctly first try.
Replies
I agree about Obsidian Sync, I'm a happy user.
A distinction worth making is between "self-hosting" (running docker-compose, Proxmox, etc.) and "local-first software" (applications that store data on your own machine with no cloud infrastructure required). The former is hard, the latter is just how desktop software worked before SaaS took over.
In small business software the shift has been nearly total. Tools aimed at craft makers, small food producers, etc. have almost universally migrated to monthly subscriptions. The practical result: you're paying $tens-$hundreds/month to track whether you have enough beeswax for your next candle batch, the price increases annually, and if the vendor folds you get 90 days to export your data (if you're lucky).
These users won't set up a homelab, but a desktop app that installs normally, stores data locally, works offline, and has a one-time price is achievable - I've been building one [1] and it's a reasonable middle ground between "trust us with your data forever" and "configure your own NAS."
[1] https://kitted.site (inventory and production management for small manufacturers)
> The only good implementation I know of personally is Obsidian Sync
Obsidian Sync gets around a major platform problem with Apple iOS devices, which is that they don't allow one app to change the data of another. You can use Syncthing for local E2EE sync, but it won't work on your mobile due to this. It works fantastic machine to machine. I'm paying for Obsidian Sync now just to get around that, but it shows how some of the platforms are made to prevent functionality. Ostensibly its for security, but I'd argue the benefit is mainly financial for app makers (and therefore their app store cut).
> half the world couldn't tell you what a terminal is and they're focused on other things in life instead of learning how computers work.
Thankfully, the converse: the computers these days are focused on nothing else but learning how humans work.
Hell, half the world doesn't even have a computer with a physical keyboard.
> If we are to solve this problem, we're going to have to make self-hosting easy enough
It used to be easy enough in the 90s, when plenty of folks had their own custom website. You signed up with a hosting company; they provided you with a bunch of different ways to upload files; your website was hosted.
Of late, I've decided that the problem is that HTML development halted at what is still a very beta product. HTMX is a reasonable attempt to continue the spirit of HTML; where I'm going with this is that I think HTML should have continued to add enough reasonable features, without needing Javascript or massive amounts of CSS, so that most websites most people wanted to develop would still be straightforward enough to do. HTML stopped before it even had a usable <table> with sorting and filtering defined, so we've spent decades with inferior tables in every web app that all suck compared to what we got used to with even Windows 3.1 apps. HTML finally grew date and colour pickers but it should have had all kinds of rich UI controls and behaviour that would have made it totally unnecessary for people to build all the Javascript middleware that essentially treats the browser as a display canvas and otherwise totally reimplements the GUI from scratch. And we wonder why the beautiful new Macbook Neo is kneecapped by only having 8GB????
It's time for HTML6. My standard will be: everything a restaurant website needs should be basically batteries included, with the exception of an e-commerce backend. It should all be doable in static HTML files with almost no Javascript and really just enough CSS to set artistic theming elements instead of having to do arcane shit just to position things.