As the "Administration" tag for my blog will attest, the sysadmin side does take a lot of work. Every now and then something is down, something needs defending (like the current HN users all clicking on the site which is running without any caches…), and on and on. In my experience, it's not so much cybercriminals that keep me busy but bad programmers that write spiders that treat every website in the world like corporate infrastructure. Stuff like git, cgit, radicale, a bunch of web apps, a netnews server, an IRC server, those things aren't tricky. But of course I also decided not to host mail, matrix, Next Cloud and so on. You need to pick your battles, too.
In the past I ended up in a place where I self-hosted virtually everything I consumed within my network that could be self-hosted. Right down to major OS software update caching.
It’s definitely exhausting and I hit upon a point where it felt like I couldn’t distinguish between home and business infrastructure. I expanded to include family stuff, and that became a massive headache. When you do that, it’s not always easy to extricate yourself either. It took several years to downsize so my services were only consumed by me again. Then a few more to fully move to non-self-hosted almost-everything.
I used to host my mail but that eventually becomes a headache. It didn’t become an admin headache, but overtime it became either/or (depending on the time) a technical liability or a security concern.
These days I try to buy/pay for the services I consume instead of self-hosting. However I’ve been moving back to assuming a “self hosting first” mindset because it seems like so many tech companies are actively trying to alienate their user base lately.
Tailscale does make sharing services amongst groups of technically minded friends _crazy_ simple.
I set up a net news server, created two groups, hooked up to another net news server, we started peering… It sure is possible! And with peering, newsgroups are federated. They will survive individual servers going down.
For some long term perspective, I liked Kris' blog post "On cache problems, and what they mean for the future" ending with the following passage:
"That challenges the fundamental abstraction of Unix, though, because in Unix everything is a file, which is a linear array of bytes, and is being accessed through a file handle. Now, with Optane persistent data may be no longer behind a file handle, but a special kind of memory, and data does not have to be crystallized into serialized structures before persistence. In fact, the memory may be so fast that we might not have time to do that.
We require a different compute abstraction instead. Which means, when we have it, the result will finally, after five decades, not really Unix any more."
http://blog.koehntopp.info/index.php/1960-on-cache-problems-...
