Files
rose-ash/next
giles 897449cb35
Some checks failed
Test, Build, and Deploy / test-build-deploy (push) Failing after 21s
fed-sx-m1: Step 3c.a segment rotation — log:open_disk/3, <ActorId>-NNNNNN.log filename, threshold-driven rotation; 10/10 log_rotate tests
`next/kernel/log.erl` rewritten around a `seg_lens :: [N0, N1, ...]` per-segment entry-count list + a `seg_size` byte threshold. Filename
scheme moved from `<ActorId>.log` to `<ActorId>-NNNNNN.log` (6-digit zero-padded) so `file:list_dir`'s alphabetical sort coincides
with numeric order.

`open_disk/3(ActorId, BasePath, [{segment_size, N}])` opts a caller into a smaller rotation threshold; `open_disk/2` keeps a 1 GiB
default that effectively never rotates (preserves Step 3b acceptance — log_disk.sh unchanged in behaviour).

Rotation rule in `place_append/4`: if the active segment's pre-append encoded size is already >= threshold AND it holds at least one
entry, the new activity opens a fresh segment; otherwise it extends the current active segment. A single huge entry that exceeds
the threshold stays alone — never rotated recursively.

On reopen, `load_all_segments` lists the dir, filters `<ActorId>-NNNNNN.log`, sorts numerically (insertion sort — `lists:sort/1`
isn't registered in this port, only `lists:append/2`/`lists:reverse/1`/`lists:filter/2`/etc.), reads each via `try_read_segment`,
and concatenates the entries to rebuild flat `entries` + `seg_lens`.

Erlang-port gotchas worked around during this iteration:
(a) String literals like `"foo"` in this port are NOT charlists — `[H|T] = "foo"` badmatches and `length("foo")` errors as "not a
    proper list". `parse_segment_name` builds prefix/suffix from `atom_to_list/1` + explicit `[$-]` / `[$., $l, $o, $g]` cons.
(b) Cross-arg variable repetition (`strip_prefix([C | Rest], [C | PRest])`) was rewritten to explicit `case C =:= P` for robustness.
(c) `Pattern = Binding` syntax in a case clause (`[_|_] = Lst when length(Lst) > 1 -> ...`) errors as "unsupported pattern type
    'match'" — replaced with `Lst when is_list(Lst), length(Lst) > 1`.

Tests:
- new `next/tests/log_rotate.sh` (10 cases): no-opt single-seg-after-3, rotation-fires-on-threshold, rotated-chronological,
  reopen-rebuilds-history, reopen-rebuilds-same-seg-shape, huge-single-entry-stays-1-seg, append-after-huge-keeps-order,
  tip-monotonic-across-rotations.
- `next/tests/log_disk.sh` updated to the new filename (`corrupted-000000.log`); stays 12/12.
- Erlang conformance 761/761 unchanged (log.erl is in next/, not lib/erlang/).

3c.a ticked in plans/fed-sx-milestone-1.md; 3c.b (gen_server-mediated concurrent appends) is the next iteration.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-06-05 07:40:48 +00:00
..

next — fed-sx Milestone 1 kernel

Single-instance, single-actor fed-sx server built as Erlang-on-SX modules. See plans/fed-sx-design.md for the architecture and plans/fed-sx-milestone-1.md for the build plan + per-step progress log.

Status

Both Step 9 smoke proof points are functional in-process:

  • 9a-pure (verb extensibility)Create{DefineActivity{Pin}} registers Pin at runtime; subsequent Pin{path, cid} activities fold into a pin-state projection. Zero kernel code between definition and use. See next/tests/smoke_pin_pure.sh.
  • 9b-pure (reactive application) — A trigger projection matches Notes tagged smoketest and derives a TestEcho carrying the source CID. See next/tests/smoke_app_pure.sh.

The remaining 9a-tcp / 9b-tcp deliverables layer TCP transport on top — see Substrate gaps below.

Layout

next/
├── kernel/      Erlang-on-SX kernel modules (.erl)
├── genesis/     SX source files for the bootstrap bundle
├── tests/       Bash test scripts driving sx_server.exe via the epoch protocol
└── data/        Runtime state — gitignored

Module map

Module Role
nx_cid.erl Canonical CID wrapper around the host cid:to_string BIF
envelope.erl Activity envelope shape, canonical bytes, time-aware sig verify
log.erl Per-actor in-memory append log (open / append / tip / replay / entries)
registry.erl Pure-functional + gen_server-wrapped registry keyed by Kind
pipeline.erl Validation driver + stage_envelope/signature/replay/schema
projection.erl Pure projection driver + gen_server-per-projection wrapper
outbox.erl Envelope construct + sign + publish orchestrator + broadcast
bootstrap.erl Genesis read/build/verify/load + one-call start/3 kernel bring-up
define_registry.erl Meta-projection fold for Create{Define*} → registry
sandbox.erl eval_pure/2,3 try/catch envelope for projection folds
nx_kernel.erl Long-lived runtime orchestrator (state + gen_server)
http_server.erl route/1,2 + format-aware GET + POST + Accept header content negotiation

Genesis bundle

next/genesis/ contains 31 SX files across 7 sections, all consumed as data (read + serialised by bootstrap:populate_registry, not eval'd):

  • 3 activity-types — Create, Update, Delete
  • 10 object-types — SXArtifact, Note, Tombstone, 6 Define* meta-types, Snapshot
  • 7 projections — activity-log, by-type, by-actor, by-object, actor-state, define-registry, audience-graph
  • 3 validators — envelope-shape, signature, type-schema
  • 3 codecs — dag-cbor, raw, dag-json
  • 2 sig-suites — rsa-sha256-2018, ed25519-2020
  • 3 audience predicates — Public, Followers, Direct

manifest.sx is the bundle root, listed in dependency-friendly order.

Tests

43 test suites, ~560+ assertions. Each script drives sx_server.exe via the epoch protocol — loads the Erlang substrate, loads relevant kernel modules via code:load_binary / erlang-load-module, then exercises behaviour through erlang-eval-ast.

Conventions:

  • Scripts marked _pure.sh exercise pure-functional state.
  • Scripts marked _server.sh (or no suffix) exercise gen_server APIs and must inline start_link with operations — the Erlang-on-SX scheduler doesn't preserve spawned processes across separate erlang-eval-ast invocations.
  • smoke_*_pure.sh are end-to-end smoke tests demonstrating the §Step 9 proof points without TCP / curl / JSON.

The Erlang-on-SX conformance gate (bash lib/erlang/conformance.sh, 729 / 729) is the no-regression contract — every commit on loops/fed-sx-m1 preserves it.

Substrate

Each .erl source file is hot-loaded at boot via code:load_binary(Mod, Filename, SourceString) (Phase 7 BIF). Tests drive the runtime via the epoch protocol:

printf '(epoch 1)\n(load "lib/erlang/runtime.sx")\n(epoch 2)\n<test-expr>\n' \
  | hosts/ocaml/_build/default/bin/sx_server.exe

The kernel calls into these host primitives: crypto:hash/2, cid:from_bytes/1, cid:to_string/1, file:read_file/1, file:write_file/2, file:delete/1, file:list_dir/1, code:load_binary/3, plus http:listen/2 (the briefing's allowed scope exception, added to lib/erlang/runtime.sx).

Substrate gaps (parked work)

These three gaps block the remaining unchecked deliverables:

  1. Term codec (3b/3c) — all three substrate fixes done 2026-06-05: erlang:binary_to_list/1 and erlang:list_to_binary/1 registered in lib/erlang/runtime.sx (iolist-aware); the tokenizer's $X branch emits the decimal char code; atom_to_list/1 and integer_to_list/1 now return Erlang charlists (standard Erlang semantics) with list_to_atom/ list_to_integer accepting both charlists and SX strings for back-compat. 759/759 conformance. The full term-codec primitive set is in place — Step 3b on-disk segment writer can encode arbitrary Erlang activity terms (atoms, ints, binaries, tuples, lists) into byte sequences using only Erlang-native primitives.

  2. SX-source eval bridge — There's no BIF that lets Erlang call into the SX evaluator on a parsed source string. Blocks evaluating the :schema / :fold / :predicate / :verify bodies from the genesis bundle. Erlang-fun stand-ins (pipeline:stage_schema, define_registry:fold, etc.) prove the API shapes; the bridge would let bundle bodies dispatch through them unchanged.

  3. Dict ↔ proplist marshalling for http:listen/2 — The native http-listen primitive calls the handler with an SX dict; the BIF wrapper's bridge would need to marshal that to / from an Erlang proplist. Blocks Step 8b-start (actual TCP listening with working route dispatch). The briefing allowed the BIF wrapper as a single scope exception; further in-place modifications need agent approval.

Bringing up the kernel

For tests, bootstrap:start/3(ActorId, KeySpec, ActorState) is the one-call boot:

KM = <<1,2,3,4>>,
KS = [{key_id, k1}, {algorithm, ed25519}, {value, KM}],
AS = [{public_keys, [[{id, k1}, {created, 0}, {value, KM}]]}],
Pid = bootstrap:start(alice, KS, AS),
%% nx_kernel + registry populated; you now have a kernel.

The HTTP layer (http_server) and nx_kernel:publish/1 flow through the same in-process gen_servers; http_publish_fold.sh is the end-to-end proof the chain works.

What's next (when work resumes)

In priority order:

  1. 8b-bridge — extend er-bif-http-listen with dict ↔ proplist marshalling so requests reach route/1 shaped correctly.
  2. 8b-starthttp_server:start/1 spawns a process hosting http:listen/2.
  3. 9a-tcp / 9b-tcp — replace the in-process smoke scripts with curl-driven versions hitting the running server.
  4. Term codec / on-disk log — needs either a new BIF or a temp-file workaround; current in-memory log keeps everything functional otherwise.
  5. SX-source eval bridge — unlocks real :schema / :fold body evaluation from the genesis bundle.