0001 — Prefer UK and EU providers over US ones for infrastructure and services

Status: accepted Date: 2026-10-02 Story: none — captured from the roadmap line "Cross-project ADRs"

Context

Domains, DNS, static hosting, source forges, email and cloud compute are bought from a dozen or so companies. Left to default, the set drifted to US firms (Namecheap, AWS Route 53 and S3, Netlify, Cloudflare, GitHub, afraid.org) because each was the cheapest or best-known at the moment of purchase. Nothing was decided; it accumulated.

Reasons that now make jurisdiction a first-class criterion, not a tie-break:

  • Legal reach. Data held by a US company is reachable under US law (CLOUD Act) wherever the region is. A UK or EU company is answerable to UK/EU law and GDPR directly, which is also where I live and where my obligations are.
  • Policy volatility. US trade and sanctions policy, and the terms of US platforms, have changed abruptly and without notice. A single foreign jurisdiction able to switch off domains, hosting and code at once is a concentration risk.
  • Where the money goes. Spend is small but recurring. Spending it with UK and EU businesses keeps a local ecosystem of providers viable, which is the only way there is a choice next time.
  • Currency and billing. GBP or EUR billing avoids FX fees and USD price drift; UK VAT handling is simpler with a UK supplier.

Constraints that cut the other way: some UK/EU providers lack an API or anycast DNS; some are dearer; several existing registrations run to 2030, so there is a switching cost and no urgency. The precedent already set: Forgejo on a Hetzner (DE) server is the primary forge with GitHub a mirror only, and static docs moved from AWS to that server.

Options

Option Complexity Testability Domain clarity Operational risk
A — cheapest or most capable, jurisdiction ignored (status quo) none — muddled: no stated rule, so every purchase is re-argued concentration in one foreign jurisdiction; terms can change under us
B — prefer UK, then EU/EEA/CH, then other non-US, then US; move incumbents at natural points low: one rule applied at purchase and renewal the inventory page shows jurisdiction per provider, so drift is visible clear: a ranked preference with a stated exception path some providers dearer or less featured; migrations spread over years
C — UK/EU only, migrate everything now high: forced migrations of working services as B clear but brittle: no exception path when no UK/EU equivalent exists breakage from migrations done for their own sake; some services have no UK/EU equivalent yet

Decision

B. New purchases default to a UK provider, then EU/EEA/Swiss, then other non-US, then US. An incumbent US service is moved when it is next touched for another reason or comes up for renewal, never purely to satisfy this rule. A US provider is acceptable when no UK/EU provider offers the capability needed, and that exception is written against the service in the domain and site inventory so it is revisited rather than forgotten. Price alone is not an exception: a modest premium over the US option is the cost of the decision and is accepted.

Consequences

  • Easier: a purchase decision has a default; the inventory page states each provider's jurisdiction, so the gap between rule and reality is one table away.
  • Harder: fewer candidates per category; registrar DNS may be slower or lack an API; some free US tiers (Cloudflare, Netlify) have no free UK/EU equivalent and will cost money.
  • Given up: the lowest .com renewal price; unlimited free edge services; the convenience of one US account that does everything.
  • Open: a UK/EU candidate list per category is kept on the family site's domains page (LAN) alongside measured DNS latency and renewal prices; this ADR holds the rule, not the shopping list.