Skip to content

docs: rename the spec to Protocol.md, drop the raw research files - #26

Merged
bennischober merged 1 commit into
mainfrom
docs/spec-restructure
Oct 24, 2026
Merged

bennischober merged 1 commit into
mainfrom
docs/spec-restructure

Conversation

@bennischober

Copy link
Copy Markdown
Owner

What

The three reviewed documentation-structure changes:

  1. Spec renamed: concept/Protocol v0.1.md → concept/Protocol.md. The old filename carried the first release's version; the file is the living, version-tracking spec (currently v4) — the name lied on every citation. The MCP-style timestamped-snapshot layout was considered and rejected (a living spec wire-referenced from ~15 places would fork every reference on each change; git history already preserves past states). All 15 references updated; the spec header notes the old name.
  2. Raw research files removed: the two Gemini deep-research documents (470 lines) are gone from HEAD — the digest is the working copy (F1–F4 closed in it), the raws were provenance, and git history (tracked since a16b4e9) preserves them byte-for-byte. The digest's provenance header is rewritten accordingly; AGENTS.md no longer sends agents to read the raw files.
  3. The prolly-tree row closed: the digest still presented it as a pending follow-up; in fact D17 rejected the mechanism (the structured version of the rejected "mega-manifest": collapses per-file identity, recreates the large-tree memory problem) while per-file conversations adopted the spirit. The row now records this with the revisit condition (spec §14 extension point, high-RTT huge-tree sessions only).

Verification

  • grep clean: no dangling old-name or removed-file references outside the two intentional provenance notes.
  • All concept/ links resolve (link + backtick forms checked).
  • Full gates green: fmt, clippy -D warnings, all suites (43 unit + loopback 23 + tree 28 + tls/quic/soak/integration).

…2-digest closure)

Three reviewed changes:

1. The spec file 'Protocol v0.1.md' is renamed to 'Protocol.md'. The
  old name carried the first release's version while the file itself
  is the living, version-tracking spec (currently v4) — the name lied
  on every citation. The MCP-style timestamped-snapshot layout was
  considered and rejected: ffsync's spec is one living document
  wire-referenced from ~15 places; timestamping would fork every
  reference at each change, and git history already preserves past
  states. All 15 references updated (5 module docs, README x2,
  AGENTS.md, CONTRIBUTING.md, the decisions ledger x3, the workflow
  doc); the spec header notes the old name for provenance.

2. The two raw deep-research files ('Designing a High-Performance…'
  and 'P2P Connection Challenge Resolution.md') are removed. The
  digest is the working copy (F1-F4 all closed with updates recorded
  in it); the raw files were provenance — tracked since a16b4e9, so
  git history preserves them byte-for-byte. The digest's provenance
  header is rewritten to say so. AGENTS.md no longer instructs agents
  to read 470 lines of raw Gemini research before making changes.

3. The digest's prolly-tree row is closed honestly: the full
  Merkle-divergence index was rejected by D17 (the structured version
  of the rejected 'mega-manifest': collapses per-file identity,
  recreates the large-tree memory problem, adds a persistent per-tree
  index) while per-file conversations adopted the spirit (the
  per-file root-hash fast path localizes divergence). The row had
  still read as a pending follow-up — a documentation bug. Revisit as
  a spec §14 extension point only if high-RTT huge-tree sessions
  dominate.

Verified: no dangling references (grep clean); all concept/ links
resolve (link + backtick checked); full gates green.
@bennischober
bennischober merged commit 260e543 into main Oct 24, 2026
4 checks passed
@bennischober
bennischober deleted the docs/spec-restructure branch October 24, 2026 13:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant