DECISION: repo-batcher do-or-delete (Item 7c) — tracked non-stripped ELF demo binary in build/

[hyperpolymath]

RESOLVED 2026-05-18 — decision = DO, not DELETE. Owner chose: reposystem is canonical for repo-batcher; keep it and port the V layer to Zig (the "DELETE … Recommended" framing below is the original pre-decision analysis, retained for history — it does not reflect the outcome). Binary untracked: PR #55 (MERGED). V→Zig port: #56. dev-ecosystem dup removal: developer-ecosystem#70. See the decision comment. Issue stays open until #56 lands.

---

Routed from Estate gitignore audit (Item 83)

The Item 83 estate .gitignore/committed-artefact audit (reverse pass 2026-05-17) found the artefact class largely already closed by the Day-3 template hardening (rsr-template-repo#43). One genuine committed artefact in reposystem was deliberately routed, not stripped, because it overlaps another tech-debt item:

• scaffoldia/repo-batcher/build/repo-batcher-demo — ELF 64-bit LSB executable, x86-64, not stripped, with debug_info, ~558 KB. Tracked since 4da0a19 (2026-05-16). The repo-root .gitignore already ignores build/ / _build/, so this is a pre-ignore committed binary — but it must not be stripped independently of the Item 7c decision below.

This is the analogue of the hyperpolymath-archive .ttc/.ttm residual, which was resolved as a standalone decision (standards#121 → hyperpolymath-archive#11, merged). With that resolved and the echo-types Agda _build/ follow-up verified clean (no tracked build output; .gitignore covers *.agdai/.agda/), this binary is the last open Item-83 residual.

Why this is an Item 7c (do-or-delete) decision, not a strip

scaffoldia/repo-batcher/ is an ATS2 + V subtree (src/ats2/…, V CLI per IMPLEMENTATION-STATUS.md). V is banned estate-wide, so the subtree cannot survive as-is regardless of the binary. Stripping just build/repo-batcher-demo would leave the larger banned-language do-or-delete question unresolved and could conflict with reposystem#33 (which already lists scaffoldia/repo-batcher examples as "likely delete, not port" in the ReScript→AffineScript migration).

Decision required from @hyperpolymath (do-or-delete)

• DELETE — drop the scaffoldia/repo-batcher/ subtree (V-banned; superseded). Single cleanup PR; history preserves the bytes. Recommended unless the subtree is being kept for a reason.
• DO — keep/rework repo-batcher off V. Then: git rm --cached scaffoldia/repo-batcher/build/repo-batcher-demo (gitignore already in place), regenerate-on-demand; no binary in tree.

Either way the tracked binary leaves the worktree as a consequence of the 7c call — not via an independent Item-83 strip.

Refs standards#66 (estate gitignore-class parent), reposystem#33 (repo-batcher migration overlap). No joint-close.

[hyperpolymath]

DECISION (owner, 2026-05-18): DO — reposystem is canonical; keep + port V→Zig

repo-batcher was found duplicated (reposystem/scaffoldia/repo-batcher/ + developer-ecosystem/scaffoldia/repo-batcher/) and absent from ambientops. Owner call: reposystem stays canonical, repo-batcher is kept (not deleted), the V CLI/FFI layer is ported to Zig, and the developer-ecosystem duplicate is removed.

Consequences, routed:

1. Item-83 binary residual — DONE. build/repo-batcher-demo untracked via PR chore: untrack repo-batcher-demo ELF binary (Item 83 residual) #55 (pre-ignore artifact; scaffoldia/.gitignore already covers build/; no gitignore change). This closes the Estate gitignore-audit (Item 83) residual.
2. V→Zig port — tracked separately (substantial; ATS2 core stays, V CLI + ATS2-V FFI bridge → Zig per estate language architecture). New issue filed.
3. developer-ecosystem dup removal — tracked separately (cross-repo dedupe; filed in developer-ecosystem).

No joint-close. #53 stays open until the port lands.

[hyperpolymath]

Progress: (1) binary untracked — PR #55 MERGED; (2) doc-debt cleared — PR #57 MERGED (#56 checklist); (3) developer-ecosystem dup removed — developer-ecosystem#70/#71 MERGED. reposystem is now the sole canonical copy. Remaining open work under this decision: the V→Zig code port (#56). #53 stays open until that lands.

[hyperpolymath]

Correction (2026-05-18): the "DO — port V→Zig" decision rested on a false premise. Verification for #56 found there is no V in source and never was (git log --all --diff-filter=A -- src/v/* empty); the ffi/zig/ layer is an unimplemented RSR template stub. So the decision's intent stands (reposystem canonical; tool kept) but the work is greenfield Zig FFI/CLI implementation, not a port — #56 retitled/re-scoped accordingly. Docs corrected: PR #57 then PR #58 (verified-accurate, supersedes #57). Net state under this decision: binary untracked (#55 ✅), dup removed (developer-ecosystem#70 ✅), docs truthful (#58 ✅), greenfield Zig impl outstanding + build-blocked on absent ATS2 toolchain (#56). #53 stays open until #56.

[hyperpolymath]

Closing: exit condition met. The decision (RESOLVED 2026-05-18 = DO, not DELETE) stated this issue stays open until the V→Zig port (#56) landed. reposystem#56 is now CLOSED (Zig FFI/CLI implemented; binary untracked via #55). Item 7c is fully discharged.