The full sequence, with what each phase delivers, is in
docs/specs/openscion-architecture.md.
This page tracks where things stand.
- Workspace, lints, CI (format, clippy, build, test, per-file coverage, rustdoc, MSRV, supply chain), and a tag-and-release workflow.
- Draft specs: architecture, data model, and API.
- Core data model (plan): object ids, paths, trees,
changesets, the in-memory log and compare-and-swap refs, root-tree
derivation by path copying; the server's object, changeset, and ref routes;
opnscn import,commit, andclone(including one-directory clones); and an OpenScion backend inopenscion-bench.
- Derived data and queries. Generation numbers, skip lists, path history,
Log,Diff,MergeBase,IsAncestor,ListTree,Stat. - Server-side rebase and landing.
Rebase,Land,Revert, file-level conflicts.
Persistent stores, the opnscn client and node daemon, git ingestion, network
tiers, line-level blame and merge, virtual checkouts, read access control.
- Replacing git for ordinary repositories.
- Offline-first operation and branch-heavy workflows.
- Code review, code search, and CI themselves. OpenScion provides the APIs they need.