[Calgooon]

The word. The owner, 2026-10-27: "arc is old af... we use arcade now for everyting". The keeper's reading (#300): GorillaPool's ARC is a rung of no gate and no courier. Arcade is the default mode. Classic ARC stays only as the hermetic stub's explicit mode until #299. Main's half landed in #302 and closed #300; this PR is beta's half.

Main into beta (98d2972). This is the first true merge of main into beta (merge base 3d195c2). Main's 0.18.x patch commits were already carried on beta, so every conflict keeps beta's side except where main brings content beta lacks:

• the Nightly on dispatch only (nightly.yml, main's);
• one ci.yml comment line on the macOS build leaving the required set;
• the HANDOFF addenda (merged clean).

Conflicted and kept on beta's side: Cargo.toml and Cargo.lock (the beta.1 pins), CHANGELOG, README, RELEASING, CLAUDE.md, explorers.rs, test_support.rs, cleanup_abandoned.rs, receive.rs, handlers.rs and its tests (#157's bound is beta's validate_internalize_instructions), and respend_witness/main.rs.

#300's files kept beta's structure in the merge. Main's behaviour follows in d87713f through the typed config (A7: no env read outside src/config.rs). Beta then moved (#138); it merged clean in 8075398.

The change.

• d87713f, services:
  • ServicesSettings: one arc_mode: ArcMode (src/config/arc_mode.rs) replaces arc_mode_arcade and arcade_flag.
  • How the mode is read: ARC_MODE wins when set (blank is unset) and takes arcade or classic in any case; any other value is a ConfigError naming ARC_MODE. Under Config follow-ups from PR 169's review: --json nulls and booleans, services_env's remaining parsers, the untested groups #172's rule, config prints the table with the row marked and exits 1, init runs, and every other command refuses at start. When ARC_MODE is unset, ARCADE decides: 1/true/yes is Arcade, 0/false/no is classic, anything else is refused by name. With both unset the mode is Arcade.
  • bsv-wallet config: prints ARC_MODE as arcade (default) or classic, with its source (env when ARC_MODE or ARCADE is set, else default). The log line quotes a value that contains a space.
  • services_env, the same effect as main's env read:
    • arc_gorillapool_url = None in both modes; ARC_GORILLAPOOL_URL stays the verifier's third-party store.
    • In Arcade mode there is no arc_config, and a set TAAL key is warned once as unused (toolbox docs: NORTH-STAR points at the release milestones, not the Phase ones (the hygiene pass, #316) #317).
    • arcade_url_for refuses a testnet wallet in Arcade mode without ARC_URL by name; the verifier is then off.
    • post_beef_ladder is logged once at the first services build. It lives in broadcast_verify.rs because A3's allowlist only shrinks.
  • Tests: main's tests are ported to the typed form.
• df9677c, e2e and docs:
  • hermetic.js forces ARC_MODE: 'classic', with the reason and Hermetic e2e: the stub plays Arcade V2 (the EF submit, GET /tx/{txid}, the SSE stream and the callback), so the suite runs in the default Arcade mode and classic ARC mode leaves the CLI #299 named; ARCADE stays dropped.
  • E2E-HERMETIC's rows (the /arc table, the /arc routes, the Arcade limit, the forced environment) say this.
  • README's broadcaster table, CLAUDE.md, src/CLAUDE.md, and changelog.d/300.md.
  • tools/differential runs the suite through the same applyEnv on both sides, so it needs no change.

Proof (foreground, private target dir, toolchain 1.98.1 for the ratchet and the differential):

cargo fmt --all --check                      -> clean
cargo clippy --all-targets --locked          -> Finished, no warning
cargo test --locked                          -> 29 binaries, 866 passed, 0 failed (after the #138 merge)
config / services_env / verifier tests       -> 15 passed (config::arc_mode, services_env::arc_mode_tests, the testnet verifier)
arch rules (A3, A7, A1)                      -> arch_explorer_sites, arch_config_env, arch_env_reads: ok
cargo test --test wire --test wire_consumers --test brc100 -> 18 / 1 / 5 passed (the corpus byte for byte)
cargo test --test gate_vectors -- --nocapture -> gate vectors: 0 divergences, every one recorded: true
cargo deny --locked check advisories         -> advisories ok

The wire worlds build their own mock services and read only quiet_config(), whose mode is now Arcade by default. No fixture reads the broadcaster config, and the corpus replays byte for byte.

./target/release/bsv-wallet config, scratch wallet, CHAINTRACKS_URL=off:

no ARC env:            ARC_MODE   arcade (default)   default
                       ARCADE     -                  default
ARC_MODE=classic:      ARC_MODE   classic            env
  (TAAL_API_KEY set)   TAAL_API_KEY  <set>           env      (the value printed 0 times)
ARC_MODE=bogus config: ARC_MODE   bogus   env  REFUSED: not a broadcaster mode: `arcade` (the default) or `classic` (deprecated, the hermetic stub's until #299)
                       Error: ARC_MODE is "bogus", which is not a broadcaster mode: ... (exit 1, the table printed)
ARC_MODE=bogus balance: the same Error, exit 1 (refused before any service is built)
--testnet, no ARC_URL: ARC_URL    unset              default

Hermetic, E2E_HERMETIC=1 node tests/e2e/run.js (Node 24, binaries from the lane's target dir):

  Passed:        24/24
  Walled:        168 attempt(s) to leave the machine: 168 known (CONNECT api.whatsonchain.com:443, CONNECT api.bitails.io:443), 0 new
  Unserved:      0
stub log: 84 arc POST /v1/tx, 42 arc GET /v1/tx/<txid>, no Arcade route (classic ARC posts only)

The base beta binary under the differential walls the same 168 with the same 84 classic posts. The walls come from the toolbox's PromiseAll Bitails and WhatsOnChain post rungs, so they predate this PR. E2E-HERMETIC's "walled 0" sentence (line 193) is stale on beta.

Differential (tools/differential --base origin/beta, and the floor):

vs beta:                         DIFFERENTIAL: 31 differences over 359 calls and 35 commands
  every one a CLI command's .stderr: the new `post_beef_ladder` info line (mode="classic",
  providers=TaalArcBeef,Bitails,WhatsOnChain), main's #300 behaviour, logged once per process
vs beta, that line removed:      DIFFERENTIAL: 0 differences over 359 calls and 35 commands
floor (--same on the head):      DIFFERENTIAL: 0 differences over 359 calls and 35 commands

Python tools and checks:

python3 -B tools/test_differential.py        -> Ran 49 tests, OK (skipped=1)
python3 -B tools/changelog --check --base origin/beta -> changelog: added fragment: changelog.d/300.md
python3 -B tools/test_scoreboard.py          -> Ran 89 tests, OK
scoreboard --check --base (base tool on the base tree) -> no number moved the wrong way from the base (35 values)
python3 tools/check-links                    -> 25 files, 0 broken links
python3 -B tools/check-cites                 -> 266 citations, 0 findings

The ratchet, last, after git fetch and the #138 merge (RUSTUP_TOOLCHAIN=1.98.1 python3 tools/ratchet --check --base <(git show origin/beta:docs/ratchets.json)):

  clippy_pedantic 825, unwrap_expect 6, files_over_800 10, tests_files_over_800 3,
  functions_over_60 86, cognitive_complexity 19, duplicated_tokens 32570 (record 32572), missing_docs 148
ratchet: no number above the record or the base record

Not this PR's: the stub's Arcade face (#299), the toolbox's TAAL provider behind Arcade (toolbox #317), and the mainnet gate. docs/NIGHTLY.md:3 still describes the 03:00Z schedule that main's merged dispatch-only word removed.

Refs #300

🤖 Generated with Claude Code

[Calgooon]

Fix round (the cold review's findings 1 to 4)

Commits (on 8075398; nothing rewritten, a plain push):

• 7aa5bcf merge origin/beta 8644a0f into the branch (finding 3): no conflict.
• 0e339dd .env.example: the broadcaster blocks say Arcade is the default (finding 1). This commit touches only .env.example; it is the one to cherry-pick to main.
• 15893e9 docs/NIGHTLY.md: dispatch only, in the owner's words (finding 2). Line 3's cron sentences and the concurrency paragraph's "the schedule runs from main" are gone; PR 308's table is unchanged. RELEASING names no dispatch on beta or main yet, so the page does not claim it does (The owner's word (2026-10-27): the Nightly is a waste of resources; its schedule goes, the measurement runs on dispatch before a minor and on demand; v0.19.0-beta.1 reads the required set #303's beta docs lane).
• a95f83d docs/E2E-HERMETIC.md: the run walls 168 known attempts, not 0 (finding 4), from this run's numbers.

Proof (foreground, CARGO_TARGET_DIR=/Volumes/Lanes/targets/p1-300-arcade-beta-fix, Node 24.11.1):

cargo fmt --all --check                  -> exit 0
cargo clippy --all-targets --locked      -> Finished, 0 warnings
cargo test --locked                      -> 29 binaries, 866 passed, 0 failed, 0 ignored (#193's test passed; no retry)
cargo test --test wire_consumers --test wire -> wire 18 passed, wire_consumers 1 passed; tests/wire's 142 fixtures (+MANIFEST) and wire_consumers' 48 identical to origin/beta
arch tests (config_env, env_reads, explorer_sites, no_error_words, no_prefix_errors, no_unwrap, spawn_sites, sql_sites) -> 8 binaries, all ok (1,1,1,2,1,6,1,1)
E2E_HERMETIC=1 node tests/e2e/run.js --hermetic:
  Passed:        24/24
  Walled:        168 attempt(s) to leave the machine: 168 known (CONNECT api.whatsonchain.com:443, CONNECT api.bitails.io:443), 0 new
    CONNECT api.bitails.io:443 x84 / CONNECT api.whatsonchain.com:443 x84
  Unserved:      0
  stub: arc POST /v1/tx 200 x84, GET /v1/tx/{id} 200 x42
python3 -B tools/differential --base origin/beta -> DIFFERENTIAL: 31 differences over 359 calls and 35 commands
  every one a CLI command's .stderr whose head differs at the `post_beef_ladder` info line (main's #300 line, logged once per process), as the PR body counts; nothing else
python3 -B tools/changelog --check --base origin/beta -> changelog: added fragment: changelog.d/300.md
python3 -B tools/test_scoreboard.py      -> Ran 89 tests, OK
scoreboard --check --base (base tool on the base tree) -> NORTH-STAR.md is current; no number moved the wrong way from the base (35 values)
python3 tools/check-links                -> 25 files, 0 broken links
python3 -B tools/check-cites             -> 266 citations, 69 line-count rows checked, 0 findings

The ratchet, last, after git fetch (beta still 8644a0f), RUSTUP_TOOLCHAIN=1.98.1 python3 tools/ratchet --check --base <(git show origin/beta:docs/ratchets.json):

  clippy_pedantic 825, unwrap_expect 6, files_over_800 10, tests_files_over_800 3,
  functions_over_60 86, cognitive_complexity 19, duplicated_tokens 32570 (record 32572), missing_docs 148
ratchet: no number above the record or the base record

Not this round's: the .env.example line 36 em dash predates this PR; cargo deny was not run in this round.

🤖 Generated with Claude Code

[Calgooon]

Landing shape (the captain). beta's protection requires a linear history, so the merge-shaped branch (main merged into beta) could not land: the rebase-merge replays main's 32 commits and conflicts where the merge had resolved them. The branch is now a linear series on beta 9daefc6 with the same content: one port commit carrying the merge commit 98d2972's content (git cherry-pick -m 1, the merge's own resolution: beta's content where the lines differ, nightly.yml main's plus PR 308's floors, ci.yml beta's plus main's one comment line), then the PR's five commits cherry-picked in order (d87713f, df9677c, 0e339dd, 15893e9, a95f83d; the trailers kept), then the re-check's one dictated docs fold (.env.example line 9, commit bef225c, under the 2026-10-26 ruling (b)). Proof: git diff between the series before the fold and the re-check's own trial merge (a95f83d merged with origin/beta, tree 942c80e7) is empty; the fold touches only .env.example. main keeps its own commits; from here a stable patch comes to beta as a cherry-pick backport, never a merge (the lanes' README gains the sentence).