Why I still self-host the boring parts
Every few months someone publishes a "stop self-hosting" post, and every few months I nod along to most of it — managed email, managed DNS, managed backups are all correct choices — and then I keep running this site off a $10 VPS anyway. Not out of ideology. The math just works out differently for the boring 90% of a personal stack.
What "the boring parts" means
The stuff with no state worth protecting and no uptime requirements anyone would pay for:
- This site — static HTML, rebuildable from a git repo in minutes
- An RSS reader's worth of feeds, fetched hourly by cron
- One SQLite database of personal miscellany
- A scratch space for scripts that need a public IP
None of it is precious. All of it would cost more attention to run on four separate managed services than on one small Linux box I already understand.
The actual cost model
The honest accounting: about an hour a month of maintenance (upgrades, log skimming, one backup restore test) and ten dollars. In exchange I get a place where experiments are free — no new service to evaluate, no pricing tier to check, no API to learn. The marginal cost of "what if I run a little thing that does X" is a config file and a cron job, not a vendor decision.
That last part is the real point. Managed services optimize for reliability, which is what you pay them for. A personal server optimizes for disposability — everything on it could burn down tonight and I'd be mildly annoyed, not harmed. That posture makes it a better sandbox than any free tier.
Where I don't self-host
To be fair to the other side: email (deliverability is a career, not a hobby), DNS (anycast exists), and anything whose failure would inconvenience people who aren't me. The line isn't "self-hosting good, cloud bad" — it's "own the parts where ownership is cheap and being a customer is expensive in attention."
For a pile of static files and one SQLite database, that line lands firmly on the side of a small, boring, well-understood box.