I mean, you could play an infinite regress game. Do you own the hardware? Do you own the cage the hardware is in? Do you own the building that the cage is in, and the land that the building is on? And then we can go toward owning the power company and the connections to anybody your servers talk to.
But in practice, self-hosting is about control. If what you're running it on is a commodity cloud instance that you could get from a half-dozen providers, then any one cloud provider has very little leverage over you.
No. If you have dedicated hardware (rented or owned) and full disk encryption you have decent control over your data. On a virtual server
you have no control and no privacy.
I think like many other terms like 'cloud', 'private', 'start up' there is a lot of gray area :). It's up to you where you draw the line. For me, self-hosting means running software in a manner where I have control of the data/application code and the server. With that definition, running software on EC2/DO/Linode is self-hosting. When I self-host using these servers, I know what the server is running and where the data resides.
Also, I think there are other similar popular terms. For those who run in their own premises, the term is on-premise. For those running it home, usually they call it home lab/NAS/home server. Self-hosting to me encompasses all this.
Also, self-hosting doesn't necessarily mean just open source. There are some amazing closed apps out there that you can self-host - emby, confluence, teamspeak to name a few.
I think it's also important to highlight that perfect is the enemy of good. I moved from Google services (A) to an open source platform... that I was paying them to host on Google Cloud (B). Sure, it's still at Google, but the ability for Google to mine data from it is significantly reduced. Of course, later, I moved to hosting it in-house (C), where I have direct physical control.
It's likely if my choices were A or C, I'd have never left A. But that B option eased the transition for me, and made it possible for me to get to the point I felt like the investment was worth it to create a fully on-premise solution.
There's a part D to this too, actually: I'm still using a service to manage the DNS and TLS for it. Eventually I should be able to move away from that too. But without the intermediate step, it'd be too prohibitive and frustrating to have moved to step C.
Great point about D. I self-host at home (it's a thinkcenter m600) but it depends on external services. I use DO DNS and Let's Encrypt TLS, backup to S3. Not to mention I rely on comcast for internet and public IP!
