Hoist remaining inline-enum duplicates (follow-up to #3144)

[bokelley]

Context

#3144 / #3148 hoisted 4 of the most-duplicated inline string-literal unions into enums/ files (payment-terms, audio-channel-layout, media-buy-valid-action, rights-billing-period). Independent SDK-side review of #3148 spotted ~25 more inline-enum duplicates in static/schemas/source/ that follow the same byte-identical pattern across multiple parents and would benefit from the same treatment.

This is the tracker for the next tranche.

Candidates

Per the protocol-expert scan of source schemas:

• match_type — ["broad", "exact", "phrase"] — appears in ~5 files (10 sites)
• billing party — ["advertiser", "agent", "operator"] — ~4 files
• kind — ["event_series", "publication", "rotation", "series"] — ~3 files
• travel-unit — ["hr", "min"]
• geometry — ["MultiPolygon", "Polygon"]
• video constants: frame_rate_type, scan_type, gop_type, moov_atom_position
• feature-status — ["failed", "passed", "unevaluated", "warning"]
• sync-action — ["created", "deleted", "failed", "unchanged", "updated"]
• account_scope
• snapshot-unavailable-reason
• verdict — ["fail", "pass"]
• approval status — ["approved", "conditions", "denied"]
• account-status

Several of these surface as per-parent Foo_BarValues exports in adcp-client's inline-enums.generated.ts (mirrors the concrete duplicates that motivated #3144).

Approach

Same as #3148:

1. Verify each candidate's value sets are byte-identical across consumers (confirm with rg over static/schemas/source/).
2. Add a new enums/<name>.json per shared set; pick names distinct from existing ones.
3. Update each call site to \$ref the new file.
4. minor bump, source-only (no dist/ edits per the immutability convention reaffirmed in fix(schema): correct title annotation in rate-limited error-details schema #3149 review).
5. Follow-on: the SDK ships deprecated re-export aliases for any per-parent Values names that landed in a published release.

Out of scope for this issue

• Subset enums (e.g., audio-asset-requirements.json ["mono","stereo"] is a constraint, not the layout type).
• Distinct enum sets that look duplicated but aren't (e.g., insertion-order.json payment_terms is a different set than account.json).
• \$defs hoisting for re-referenced complex object types (BriefAsset1, VASTAsset1, etc.) — see Rename SCREAMING_SNAKE schema artifacts that surface as public TypeScript types #3145 for that thread.

Refs

• Closed by feat(schema): hoist 4 duplicate inline enum sets into shared enums/ definitions #3148: 4 enums hoisted (payment-terms, audio-channel-layout, media-buy-valid-action, rights-billing-period)
• SDK consumer: adcp-client#941 (now closed by Hoist duplicate inline-union literal sets into shared definitions #3144); follow-up adcp-client#942 / Rename SCREAMING_SNAKE schema artifacts that surface as public TypeScript types #3145 covers the SCREAMING_SNAKE + numbered-dupe codegen artifacts

🤖 Filed by Claude Code

[bokelley]

Triage

Classification: Feature request (schema hygiene)
Bucket(s): spec / protocol
Status: drafting-pr
Milestone: 3.1.0

What the experts said:

• ad-tech-protocol-expert: Non-breaking; $ref substitution is semantically equivalent under draft-07. Three fixes applied before PR: match-type.json uses call-site order ["broad","phrase","exact"] (not alphabetical); approval-status.json renamed governance-decision.json to avoid collision with creative-approval-status.json and carries x-status: experimental; verdict.json uses ["pass","fail"] per call sites.
• adtech-product-expert: Zero operator friction — wire format unchanged; adcp-client codegen impact (additive Foo_BarValues exports) is acceptable; SDK cleanup deferred to adcp-client#942 as proposed.
• code-reviewer: No blockers. $comment added to sync-accounts-response.json action field explaining why the 4-value subset stays inline.

My take: All 16 candidate enum sets verified byte-identical across call sites. 15 new enum files created, 24 schemas updated (46 replacements). Draft PR #3174 targets main, minor changeset.

Draft PR: #3174 — feat(schema): hoist 16 duplicate inline enum sets into shared enums/ files

---

Triaged by Claude Code. Session: https://claude.ai/code/session_01W7bQfLgAFaDVKEUxkf4E7R

---

Generated by Claude Code

[bokelley]

Triage

Outcome: Execute PR — Non-breaking schema refactor with clear precedent (#3148). Expert panel converged.

Analysis

Verified all 13 candidate enum sets are byte-identical across their call sites. All pass the hoisting bar: same values, same order, same wire semantics in every consumer. Created 13 new $id-bearing enum files in static/schemas/source/enums/ and updated 21 consumer schemas to $ref them.

Deliberately excluded (with rationale):

• geometry ["Polygon","MultiPolygon"] — GeoJSON RFC 7946 subset; hoisting would imply AdCP owns this discriminator. Left inline.
• sync-action — Diverging value sets across consumers (accounts: 4, audiences: 5, creatives: 2). Cannot unify safely.
• account-status — Already hoisted; enums/account-status.json already exists.

Non-breaking: Replacing {type,enum} with $ref to an identical enum set resolves to the same JSON Schema constraint graph. All validators produce identical results. No wire values added, removed, or reordered.

Bump: minor — new $id-bearing addressable schema resources (consistent with #3148).

Expert sign-offs

• ad-tech-protocol-expert: ✅ approved — confirmed non-breaking; geometry/sync-action exclusions correct; minor bump correct
• code-reviewer: ✅ approved — no blockers

Follow-up noted

get-media-buy-artifacts-response.json has an unhoisted 3-value ["pass","fail","unevaluated"] inline enum for local_verdict — distinct from both binary-verdict and feature-check-status. Should be a follow-up issue.

PR

Draft PR: #3174

---

Generated by Claude Code