Self-Hosting vs Managed: A FOSS Perspective
The FOSS community has a self-hosting bias. If you can run it yourself, why pay someone else? This instinct is healthy — data ownership, software freedom, and technical skill-building are genuine values. But it can also lead to burnout, insecure infrastructure, and projects that spend more time on ops than on their actual mission.
The Self-Hosting Sweet Spot
Some FOSS services are genuinely low-maintenance and high-value when self-hosted:
Gitea / Forgejo — a single Go binary. Install it, configure a reverse proxy, set up backups. You’re done. Maintenance: ~15 minutes per month for updates.
Vaultwarden — Bitwarden-compatible password manager in Rust. Docker container, 50 MB RAM, set up once and forget about it. Critical to back up, but otherwise nearly maintenance-free.
Plausible / Umami — web analytics that don’t track users. Simple setup, minimal resource usage, no ongoing tuning required.
Uptime Kuma — monitoring tool that just works. Docker, simple UI, push notifications. Set it up, add your services, and it does its job.
The Self-Hosting Danger Zone
Other services are deceptively difficult:
Email (Mailcow, Mailu, mailcow) — email self-hosting is a trap. The software works, but deliverability doesn’t. Without established IP reputation, your mail lands in spam. Gmail and Microsoft actively penalize small mail servers. Unless email infrastructure is your project’s purpose, use Fastmail or Proton Mail.
Matrix — federation is complex and resource-intensive. The Synapse server is a Python monolith that consumes surprising RAM. State resolution bugs can cause rooms to break. Unless you need federation control, use a public Matrix homeserver.
Jitsi Meet — works beautifully for small calls, but scaling beyond ~20 participants requires significant infrastructure. Bandwidth costs add up quickly if usage is frequent.
The Hybrid Approach That Actually Works
The best FOSS infrastructure strategy is hybrid. Self-host what’s low-maintenance and high-value. Use managed services for what’s operationally complex.
Self-host: Gitea (code), Vaultwarden (passwords), Plausible (analytics), Woodpecker (CI/CD), Uptime Kuma (monitoring)
Managed: Email (Fastmail/Proton), video calls (use public Jitsi instances), DNS (Cloudflare or deSEC). For the production database and web tier — OpsHelp’s managed hosting is built on the exact same FOSS stack you’d self-host.
Conditional: Nextcloud (self-host if you have time; otherwise use a managed Nextcloud provider), Matrix (use a public homeserver unless you need federation control)
The Time Budget
Self-hosting costs time. Budget it honestly:
- Initial setup: 2-4 hours per service
- Monthly maintenance: 30-60 minutes per service
- Quarterly updates: 1-2 hours
- Incident response: unpredictable
If your FOSS project can absorb 4-8 hours per month of ops work, self-host the full stack. If not, be selective.
When Managed Hosting is the FOSS Choice
Managed hosting isn’t a betrayal of FOSS values — it’s resource management. Spending your limited time maintaining email servers when you could be improving your FOSS project is a worse outcome for the community.
The FOSS movement benefits from your code contributions, documentation, and community building — not from you debugging Postfix configuration at 2 AM.