Git hosting that never leaves Europe

Original Article ↗ Hacker News Discussion ↗

A new EU-based Git hosting service, Pushin.eu, aims to keep code entirely within European jurisdictions and promises not to use repositories for AI training, drawing interest from developers seeking an alternative to US‑centric platforms. Commenters welcome the performance, Elixir/Rust tech stack, and focus on commercial as well as open-source projects, but raise concerns about missing legal and pricing information, vague marketing language around “blocking slop,” and long-term trust for a one-person, invite‑only beta. Comparisons with Codeberg and self‑hosted Forgejo highlight trade-offs between ideological constraints, sovereignty, and the need for a reliable, enterprise-friendly forge in Europe.

Trust, Transparency & Maturity

  • Many commenters like the idea of EU-only hosting but see “zero trust signals”: no legal pages, company details, pricing, or clear policies at launch.
  • Founder explains the site hit HN earlier than planned; focus was on app stability and data durability rather than the marketing/legals.
  • Several argue that trust docs (ToS, privacy, ownership model) are not “marketing” but core to entrusting source code.
  • Confusion around wording like “track” vs. “store/collect” user data reinforces calls for clearer legal language.

Business Model, Pricing & Roadmap

  • Service is in invite-only beta, self-funded, with a target GA around 2027.
  • Planned GitHub/GitLab-style subscriptions for individuals and teams; exact pricing not set yet.
  • Current quotas (logs, artifacts, cache) are shared; they may change with paid tiers.
  • Users request knowing pricing before migrating serious work.

Tech Stack, Hosting & Europe/EU Scope

  • Stack: Elixir/Phoenix/LiveView for most logic, Rust + gitoxide for Git, S3-compatible storage, Postgres, bare-metal servers in a French datacenter.
  • Git implementation stores packfiles in S3 with Postgres-based locking; WAL-based designs were evaluated and rejected for latency reasons.
  • Slogan “never leaves Europe / 100% EU” prompts questions about whether that means EU only, Europe broadly, or if future non‑EU regions are possible.
  • DNS is via a provider using Cloudflare; commenters note this introduces a US dependency. Lack of IPv6 and DNSSEC is criticized.

AI, “Slop”, and Bots

  • Homepage claims “Bots are welcome. Slop gets blocked,” triggering questions.
  • Clarified intent: allow bots/LLM tooling but try to reduce low-quality, spammy PRs/issues.
  • Planned mechanisms: invite-only onboarding, reputation/vouching system, rate limits on new contributors, downranking suspected low-quality contributions while leaving final decisions to maintainers.
  • Some see this as welcome spam control; others view “slop” as vague and potentially ideological.

Positioning vs. Alternatives

  • Compared with Codeberg: that platform restricts to FOSS, bans crypto and AI-generated-heavy projects, and is ideologically opinionated; this new service targets both public and private, including commercial/enterprise use.
  • Some ask how it differs from simply running Forgejo/other forges on European hosts like Hetzner; the answer emphasizes managed hosting, community features, and EU focus.
  • A subset of users prefer self-hosting but are curious and may test later.

UX, Branding & Naming

  • Many praise the interface as extremely fast and keyboard-friendly.
  • Others criticize “vibe-coded” design, generic EU branding, and name ambiguity (“pushin” read in various unintended ways).
  • Concern that a heavily derivative visual style without strong branding or trust signals could hurt long-term adoption.

CI & Integrations

  • Basic CI already works via Gitea-compatible runners; better UI and managed runners (possibly Firecracker-based) are planned.
  • Some users want simpler, more portable CI (e.g., docker+script) and/or third‑party CI integrations.

Invites & Access

  • Registration requires invites, even for filing issues, which some find too restrictive but consistent with the anti‑slop strategy.