The other benefit is that using self hostable software makes it harder for centralised deployments from screwing users over.

It is harder (but not impossible, and not without it's own inconvenience) for mastodon.social to do a rug pull because there are near-identical alternatives that others (or yourself) host.

It's all niche stuff that only a few people use until there are watershed moments like the Twitter and Reddit fuck ups that push large swathes of users to look for an alternative. Then suddenly it's not a niche product and it's important that the kinks, bugs, and onboarding has been worked out during those years of being niche.

People are absolutely getting sick of subscriptions. It's also getting easier to self host. Tailscale has been a game changer for me personally as I just had no confidence in getting my services working correctly over the internet without getting pwned

> It's all niche stuff that only a few people use until there are watershed moments like the Twitter and Reddit fuck ups that push large swathes of users to look for an alternative.

And then after poking around for a week, they go back to Twitter and Reddit.

As something of a dumdum myself I think I know why. Corporations want people to be able to do their thing as easy as possible to make money, while people not directly motivated like that and not motivated to make it as easy as possible can do anything else. So instead of "make sure it's just a button click" it's "what, you didn't read all 580 pages of the documentation and all the changelogs and the code on github and compile it yourself on a custom built $40,000 machine? We don't help your kind around here go away" and yeah people go right back to windows or twitter or whatever.

Jokin aside I'm just trying to explain there is a real problem there. Feeling smug about the result of that problem doesn't fix it but it is really easy to do

The difference between self-hosting most things, and Twitter and Reddit (and Facebook and Slack and Discord) is the network effect. If I wanted to self-host my pictures that I share with friends, I can still just send them the URL. They might be annoyed that they're not on Instagram and have to use a web browser instead, but the people that want to see how my long weekend went will go see the pictures. To self-host something like Reddit, I need to convince other people to change their habits and their choice of platform. As not-a-million-dollar-corporation, my ability to have a polished UX is rather more limited, so I can see why someone would go back Twitter and Reddit.

I think there’s a middle ground for cooperatives but the old problem of fairness rears its ugly head. I don’t want to pay for 20% of something if I’m only getting 5% of the benefit.

If the cost to help maintain the thing is something nominal (say, $20/mo, even as much as $40/mo) to maintain, then I see it as as form of mutual aid and am happy to pay it to support my friends and friends-of-friends.

As long as one guy isn’t getting 80% of the benefit, I’m game.

The other thing to consider is that self-hosting is not a binary option - there are degrees to it. On one end, I can upload a Docker image/OCI tarball to a cloud provider and get a service up and running with plenty of application-level customization. Somewhere in the middle, I can get a private server and have a bit more low-level control over my deployments, like tweaking some sysctl parameters, or running a custom-built Linux kernel. On the other end, I can literally buy my own rack server, with all the hardware I need or want installed in it, and send it to a colo for hosting and upkeep (or build my own data center, if I have the money).

Stuffing it into your basement, which is the "build your own data center" option, isn't prohibitively expensive, not does the hardware have to be. There's a gulf of prices between a Raspberry Pi and a new Dell or HP server. On top of that, getting 5 nines of uptime is costly, but we're not trying to self host Google.com here. If my personal file server goes down, my friends'll eventually notice but we're talking about a service that gets 0 rps (requests per second) when all of us are all sleeping, so no nines is sufficient. More would be great, but like you said, it's expensive.

This is the best route, in my experience.

If you're interested in tech or gaming, you usually accumulate hardware anyways - putting the old stuff to use just makes sense in most cases.

And I actually don't really agree with the article - My issue with SaaS products is not privacy. My problems are quality and consistency. My self-hosted stuff doesn't auto-update to a version that's less capable or dumb itself down to shove users into advertising flows or "new" features they want me to use. 7

It's not about privacy - it's about having the computer serve me. It's the difference between a free "financial advisor" peddling scams vs a paid agent with fiduciary duty.

My SLA is only to myself, so it’s not so bad. That said I do host some “critical” services like my password manager, and so making a mistake that takes those down can be a pain. I figure on balance it has been worth it.

just FYI, I've been running a few services on a Hetzner VM and a few others in a box at home. With an 5 minute uptime ping (uptime robot) I am consistently getting a 3 to 4 nines uptime for the home server and that can be substantially improved if I care to put an UPS (my old one failed). As others pointed out, home ISP is often not great for self hosting due to things such as asymmetric transfer speed, periods of higher latency and short, but sometimes frequent outages.

This goes both ways. Depending on organizationally managed services results in tools that are honed for institutions. Having self-hosters in the mix generates demand for simplified products that can offer equivalent capabilities with a minimum of effort and oversight. This potentially shifts the entire tool development ecosystem toward humans instead of large groups that have effort to spare.