Skip to content

Latest commit

 

History

History
49 lines (37 loc) · 2.55 KB

File metadata and controls

49 lines (37 loc) · 2.55 KB

Providers

shipmate's job is to span host + DNS + registrar so you don't hand-click across three dashboards. This is the per-provider reality — capabilities, costs, and where the automation stops.

DigitalOcean App Platform — ✅ live

  • Cost: ~$5/mo flat (basic-xxs, always-on container). No commercial restriction — fine to run a real product in the wild.
  • Scaling: you choose the instance size / count. A single small box won't absorb an uncontrolled traffic spike — size up before a big launch.
  • Deploy: .do/app.yaml spec + doctl apps create/update, GitHub deploy-on-push.
  • Secrets: never in the committed spec. Public build-time vars can be inline; secrets go in the DO dashboard as Encrypted env vars.

Vercel — 🛠 planned

  • Cost: Hobby is free but non-commercial only. A funded/commercial product must use Pro ($20/mo) + bandwidth.
  • Scaling: serverless / edge — auto-scales under spikes with no config.
  • DX: best-in-class for Next.js (made by the Next team) — zero-config, per-PR preview URLs.

Picking between them

  • Prototype / demo: Vercel (free, best DX, shareable preview links).
  • Real product, cost-sensitive: DigitalOcean ($5 flat beats Pro's $20 and is commercial-clean).
  • Both map apex domains and subdomains equally well.

DNS — the three modes

DNS is the actual painpoint, and the fix is where DNS lives, not "automate the registrar."

  • Mode A — DNS on the provider (DO DNS / Vercel DNS / Cloudflare): fully automatic. Point the domain's nameservers at the provider once; from then on every subdomain + cert is created from the deploy spec. Zero manual steps, forever.
  • Mode B — DNS at the registrar (e.g. SquareSpace): manual once. SquareSpace has no DNS API — shipmate prints the exact CNAME, you paste it. Unavoidable until you migrate DNS off it.
  • Mode C — no custom domain yet: ship to the provider's default URL (<app>.ondigitalocean.app) and add a domain later.

⚠ Migrating an existing domain's nameservers (Mode B → A) means recreating ALL its current records (email/MX, other subdomains) on the new provider first, or those services break. A fresh dedicated domain straight onto provider DNS avoids the migration entirely.

The registrar gap (and why it's the interesting part)

PaaS platforms own the host. Nobody owns the seam across host and DNS and registrar — especially registrars with no API. That gap, plus a safety model for letting an agent touch irreversible/billable infra, is the part worth building.