content(blog): executable ontology vs knowledge ontology — post 2 of the positioning wave (slot 2026-10-20)

[objectstack-fleet]

Fixes #228

Post 2 of the positioning wave ruled on #209, slot Tuesday 2026-10-20. Merge on the slot day only: the charter says "merge time is publish time".

Re-slot. The post was slotted for Tue 10-13. Post 1 missed 10-08 and landed on 10-15 as #212, so the wave slipped one week. Commit e2a55f1 moves the post's date to 2026-10-20 and moves the sibling's updated to 2026-10-20T10:00:00+08:00 in all eight locales. 11 lines change. Gates on that commit all exit 0: content:lint, check, build (891 pages, tree clean) and seo:smoke. Built JSON-LD gives datePublished 2026-10-20 for the post and dateModified 2026-10-20T02:00Z for the sibling.

Branch. This branch holds six commits from the staged lin/issue-228, replayed onto main @ a7bd4af+1174309. The replay is patch-equal to lin/post-base..lin/issue-228, with byte-identical diffs. Head c3538ab. The staged SHAs quoted below are the pre-replay commits.

Known follow-up. A seat fixup re-dates the stack's updated lines to their real landing day, 2026-10-15, and lands later today. This PR moves the same sibling updated lines to 10-20, so after that fixup the branch takes main with a merge commit that keeps 10-20.

The post

content/blog/executable-ontology-vs-knowledge-ontology/ — Executable Ontology vs Knowledge Ontology: Why "No Inheritance, No Axioms, No Reasoner" Is the Point (the ruled title, kept whole). date: 2026-10-13, no updated, topic: app-building, audience: developer. Locales: en, zh-Hans, zh-Hant (generated by pnpm gen:zh-hant, never hand-edited).

• Angle. The second of the pillar post's three negations, argued in full: a knowledge ontology (OWL 2 / RDF) is built to be reasoned over; an executable business ontology is built to be validated, reviewed and run. The three refusals are what make a definition checkable and a diff reviewable.

• Ownable concept. A reader for every key: a key you can write has a runtime reader, or it is removed. The one figure is ADR-0049's own, quoted: the audit "found that roughly half of all spec properties are dead".

• Proof artifact (before/after, measured). The abstract key on the object schema, parsed through three published @objectstack/spec releases with the same object (scratchpad/issue-228/scripts/probe-abstract.mjs, logs probe-v15.log, probe-v16.log, probe-v161.log, probe-v17.log):

  version (npm publish date)	ObjectSchema.create()	ObjectSchema.safeParse()
  15.1.1 (2026-07-17)	accepts, abstract: true kept	accepts, abstract: true kept
  16.0.0 (2026-07-21)	throws, naming the removal	accepts, abstract silently stripped
  17.7.0 (2026-10-06)	throws, naming the removal	rejects, unrecognized_keys, with the prescription

• Honesty section. Steelmans OWL/knowledge graphs; states plainly what the executable ontology does not do (no reasoning, no open-world inference, no schema-level inheritance, no entity resolution) and that "validated" is only as true as the validators (the ledger is standing maintenance, not a finished state).

• ObjectOS appears once, as T4 (T4_STANDALONE) + E1 verbatim from src/lib/positioning.ts.

• CTA: npm create objectstack@latest my-app && cd my-app, a coding agent, npx os dev --ui. os start appears 0 times.

• No token figures. No competitor claims. OWL is cited to the W3C OWL 2 Primer by name only.

Every factual claim and its source (objectstack origin/main = 8caa131e52717ef35fe71742d758f10b8f2b77ee)

Claim in the post	Source
"no inheritance, no axioms, no reasoner — validated rather than reasoned over"; the whole "What we mean by ontology" paragraph (quoted in a blockquote)	README.md:79-89, line-exact :82-83
T1 / T2 / T3 / T4 / E1	src/lib/positioning.ts (this repo); README.md:7-16, :26-30
"OWL, RDF and knowledge graphs describe a domain, carry class inheritance and axioms, and are reasoned over…"; "The abstract key was removed from the spec for exactly that reason (ADR-0049)"	content/docs/concepts/ontology.mdx:44-48
pipeline line objectstack.config.ts -> objectstack.json -> …	ontology.mdx:61-63
"Nothing is reasoned over at runtime: what runs is what was validated."	ontology.mdx:69-70
"Lookups and master-detail links between objects" (the only relation kinds)	ontology.mdx:21
derives / enforces	ontology.mdx:65-69
os validate "rejects metadata that type-checks but would fail silently at runtime: dangling bindings, bad CEL predicates, missing security posture"	README.md:174
"an agent's errors become located, corrective text it can read and fix itself"	README.md:179
the typed gate "in the editor"	README.md:173; object.zod.ts:2646 (NoExcessObjectKeys)
audit "found that roughly half of all spec properties are dead"	docs/adr/0049-no-unenforced-security-properties.md:7
the three states, "parsed, unmarked, unenforced", "enforce-or-remove gate" (blockquote)	ADR-0049 :25-34
"worse than absent … false sense of compliance"	ADR-0049 :21-23
"Removal is preferred over marking when there is no committed roadmap…"	ADR-0049 :91-92
the repo "cites this ADR as the enforce-or-remove policy for spec-property retirement generally"	ADR-0049 :162-164
"Authoring them is a false affordance — especially dangerous for AI-authored metadata."; abstract in the verified-dead object set	objectstack#2377 (public issue body)
tombstone text (abstract was removed … Remove the key.)	packages/spec/src/data/object.zod.ts:1538-1541; :1694-1696
"tags/active/abstract REMOVED (#2377) — tombstoned in UNKNOWN_KEY_GUIDANCE"	packages/spec/liveness/README.md:958
unknown keys used to be "discarded by Zod's default .strip() … vanished with no error, no warning, and a green tsc" (#1535)	object.zod.ts:1446-1450; objectstack#1535
"Until this shape was closed these were dropped silently on the PARSE path — …"	the measured 17.7.0 message; object.zod.ts:1676-1679
changelog: "object tags/active/abstract: ObjectSchema.create() now throws a located error naming the removal. None gated anything at runtime — remove them." (16.0.0)	packages/spec/CHANGELOG.md:98572-98573 (under ## 16.0.0 at :98195)
15.1.1 declaration abstract: z.boolean().optional().default(false).describe("Is abstract base object (cannot be instantiated)")	@objectstack/spec@15.1.1 dist/data/index.js:3239 (published package)
15.1.1 Studio object form boolean field, help text "Abstract base (cannot be instantiated)"	@objectstack/spec@15.1.1 dist/data/index.js:5794
15.1.1 ledger entry (status dead, no runtime reader, authorHint)	@objectstack/spec@15.1.1 liveness/object.json:224-229
publish dates 2026-07-17 / 2026-07-21 / 2026-10-06	npm view @objectstack/spec time
the measured outcomes per version	probe-v15.log, probe-v16.log, probe-v17.log in the scratchpad
OWL 2: classes, properties, axioms (subclass, disjointness, cardinality, property chains), reasoner, open-world assumption	W3C OWL 2 Primer (linked; standard, not a competitor)

Assets (six SVGs, all pass svg:lint)

cover-en.svg / cover.svg (1200×675), two-pipelines-en.svg / two-pipelines.svg (1200×660), abstract-before-after-en.svg / abstract-before-after.svg (1200×620). English set for en; Chinese set for zh-Hans/zh-Hant. Each diagram adds information the cover does not carry (the two pipelines; the three-release timeline with ADR-0049's three states). Text fills chosen for AA on every fill they sit on: #173a36, #256a5e, #52655f, #7a5a2c; measured in Chromium with each SVG opened as its own document, lowest ratio 5.68:1 (AA needs 4.5:1).

Two build-side lessons, recorded in the commits: Astro's vendored image-size validates an SVG over its first 1,000 bytes, so a root aria-label longer than that fails astro build with NoImageMetadata — the two timeline diagrams now keep a short aria-label and carry the full description in title and desc elements; and seo:smoke judges every prose block against its nearest lang, so the two English source quotations in the zh post are blockquote elements carrying lang=en.

Also in this PR: enterprise-ontology-race-open-vs-closed (update, not retire)

Every locale (en, zh-Hans, ja, de, es, fr, ko; zh-Hant regenerated). The title, description and TL;DR no longer name the contested layer "the business semantic layer"; they name the business definition layer — house vocabulary (glossary definition-layer; AGENTS.md rule 1's "definition layer vs runtime layer"), chosen at review over "business ontology" because that collides with post 5's planned title (#231). Each locale uses its existing rendering of the term consistently in the title, the TL;DR and the race-section line (zh also its closing line): 业务定义层 · 業務定義層 · Geschäftsdefinitionsschicht · capa de definición del negocio · couche de définition du métier · 비즈니스 정의 계층. The TL;DR and race line say which of the five entrants shipped an ontology by name (Microsoft Fabric IQ, Palantir Foundry) and which a semantic layer (Snowflake, Google Looker, Databricks Unity Catalog), matching the post's own table. updated: 2026-10-13T10:00:00+08:00 on all 8 files (the publish day, per review); date and slug unchanged; tags unchanged (the post still discusses semantic layers). Descriptions kept ≤ 230 chars (en 218, de 226, es 224, fr 228).

locale	before → after (title)
en	Who Owns the Business Semantic Layer? → Who Owns the Business Definition Layer?
zh-Hans	谁拥有业务语义层 → 谁拥有业务定义层
ja	業務セマンティックレイヤーは誰が所有すべきか → 業務定義層は誰が所有すべきか
de	die semantische Geschäftsschicht → die Geschäftsdefinitionsschicht
es	la capa semántica del negocio → la capa de definición del negocio
fr	la couche sémantique du métier → la couche de définition du métier
ko	비즈니스 의미 계층 → 비즈니스 정의 계층

Gates on 835f9b5 (through scripts/pm/os-verify-lock.sh, slot www-228, lock held 141 s, waited 0 s, shared-box seconds; the same six were green on 680bfd8 before the fixup)

• pnpm content:lint → exit 0 — ✓ content lint passed (350 files, 44 glossary terms checked); advisories on the new post, verbatim: Date is in the future: Tue Oct 13 2026 (now Oct 20 after the re-slot) 00:00:00 GMT+0000 (Coordinated Universal Time) ×3 (en, zh-Hans, zh-Hant — expected for a slot-day post, date unchanged); Title is 100 columns wide (100 chars); search results show about 60, so the end is cut (the ruled title, kept); Description is 222 columns wide (222 chars); on the sibling post, verbatim: Updated is in the future: 2026-10-13 ×8 (one per locale file — expected, the publish-day value chosen at review).
• pnpm content:lint --published → exit 0 — same line.
• pnpm check → exit 0 — Result (152 files): 0 errors, 0 warnings, 0 hints.
• pnpm build → exit 0 — ✓ zh-Hant: generated 48, kept 0 hand-maintained; svg-text-fit: OK … (124 SVGs, 2661 text boxes …); [build] 891 page(s) built in 82.33s; ✓ content lint --dist passed (489 built blog pages checked); CJK line breaks, quotation marks and 94,469 internal links all pass; tree clean afterwards.
• pnpm seo:smoke → exit 0 — SEO smoke test passed (881 HTML pages checked; … 16525 prose blocks language-checked; 94459 internal links redirect-checked).
• pnpm svg:lint → exit 0 — svg-text-fit: OK.

Browser check (astro preview on dist/, Playwright 1.56.1 Chromium at /opt/pw-browsers/chromium; the full run on 680bfd8, the sibling re-take on 835f9b5)

scratchpad/issue-228/scripts/shots.mjs: the article in en and zh-Hans at 1366×625, 1440×750, 1512×830, 1920×945, 390×844 and 360×800, light and dark; the updated post in en and zh-Hans at 1440 and 390; the en blog list at 1440 and 390; each served SVG at native size as its own document. 182/182 assertions: HTTP 200, documentElement.scrollWidth equals the viewport and no element past it outside a scroll container, the cover renders, both diagrams render, no repeated image, the Updated line reads Updated Oct 7, 2026 / 更新于 2026年10月7日, the new post is the first card of the Latest section, every SVG text AA. Element crops (scripts/crops.mjs) of every table, code block, blockquote and diagram at 1440 and 360 were looked at: tables wrap at 360, code blocks scroll sideways (the #213 design), the lang="en" quotes render as ordinary blockquotes in both schemes.

Sibling re-take on 835f9b5 (scripts/sib-shots.mjs, scratchpad/issue-228/shots-sibling/): en and zh-Hans at 1440×750 and 390×844, fold and full page, 20/20 assertions — HTTP 200, no overflow, cover renders, the H1 is the new title in both languages, the Updated line reads Updated Oct 13, 2026 / 更新于 2026年10月13日 (datetime 2026-10-13); the built en page carries the new meta description and dateModified 2026-10-13T02:00:00.000Z. Looked at: all four folds and both full pages.

Acceptance notes

• src/components/ArticleList.astro:49 and :55 (recommendedLead) still say "why the business semantic layer should be open" / 「业务语义层为什么该开放」 next to the retitled Recommended card. Site code, outside this card's surface; carrier: the seat's planned Recommended-grid change (Post (slot Thu 2026-10-08): What Does "The Ontology Is the Software" Mean? An Executable Business Ontology, Defined #211 ACCEPT, answer 2). Noted, not filed.
• The phone-width body diagrams render at 326px (12px text at ~3.3px); the site-wide Phone-width reading of blog tables and diagrams: table rows run up to 2,394px on one line, and diagram text renders at 3–4px #213 case, not this post's.
• Review outcome recorded: the first staging retitled the layer "business ontology"; the coordinator ruled neither that nor "executable business ontology" (collision with Post (slot Thu 2026-10-22): Who Owns the Business Ontology When the Platform Is Hosted? ObjectStack vs ObjectOS, Stated Plainly #231) and chose the house term definition layer, and updated = the publish day; both applied in 835f9b5.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01FeA1nwBz1ohH65dvffUGKr

🤖 Generated with Claude Code

https://claude.ai/code/session_01FeA1nwBz1ohH65dvffUGKr