Maintenance, freshness, and crawler policy
This policy turns publication into a repeatable maintenance contract. It applies to every published en-US and zh-CN route, generated database snapshot, and discovery asset.
Passing automation is release evidence, not a WCAG conformance claim, hardware test, search-ranking promise, or AI-citation guarantee. Human technical and language review remains required.
Ownership and escalation
| Area | Accountable role | Release responsibility | Escalation |
|---|---|---|---|
Bilingual content and metadata | editorial | parity, terminology, sources, status | P0 coordinator |
Technical evidence | page specialist owner | current claim and test boundary | editorial |
Build and deployment | release-engineering | lockfile, CI, deploy, rollback, restore | P0 coordinator |
Availability and discovery | site-reliability | routes, links, crawl assets, stale reports | P0 coordinator |
P0 means a published path is unavailable, unsafe guidance is live, sensitive data is exposed, or a widespread destructive error is confirmed. The response SLO is within 24 hours. Role definitions and the correction intake path live in data/maintenance/owners.json.
Review SLO
| Content class | Maximum review interval | Examples |
|---|---|---|
High-change database record | 30 days | project metrics, game reports, volatile support |
Distribution, tool, compatibility, hardware | 90 days | Bazzite, Decky, device firmware/support |
Fundamentals and stable procedures | 180 days | architecture, recovery decision models |
The monthly report calculates lastReviewed + reviewAfterDays. An overdue page cannot silently remain a current recommendation. Quarterly work retests distributions and available devices; the annual hardware edition is recalculated from a fixed evidence snapshot.
Discovery and AIEO boundaries
- Each page has a unique title, description, focused editorial keywords, Open Graph description, canonical URL, reciprocal locale alternates, visible FAQ, sources, and review date.
- FAQ JSON-LD must describe the visible questions exactly. It is not advertised as rich-result eligibility.
sitemap.xml,robots.txt, andllms.txtare generated and tested.llms.txtis an additional machine-readable index, not a web standard or ranking switch.- Crawler access follows the published
robots.txt. Search and answer engines independently decide whether to crawl, index, display, or cite a page. - Short conclusions, applicability boundaries, dated evidence, tables, and rollback steps are preferred because they help both people and machine readers verify scope.
Accessibility and search quality
WCAG 2.2 is the policy reference. Automated browser and Lighthouse samples cover language, landmarks, overflow, keyboard-reachable search, and regression budgets; manual review remains necessary for meaning, reading order, focus experience, and technical accuracy.
Native xyd Orama search is evaluated in both locales with installation intent, device aliases, technical terms, a misspelling, Chinese queries, and deliberate no-result queries. Search fixtures must point to reviewed routes rather than hiding poor relevance behind a custom frontend.
Maintenance cadence
Monthly
Generate the staleness and external-link reports; triage P0 and overdue high-change pages.
Quarterly
Retest available distribution/device paths, review crawler and structured-data guidance, and exercise a preview rollback.
Annually
Recalculate the dated hardware edition, rehearse xyd upgrade and backup restoration, and archive the evidence bundle.
Frequently asked questions
Does valid schema guarantee a rich result or AI citation?Does valid schema guarantee a rich result or AI citation?
Who owns a page when several technical areas overlap?Who owns a page when several technical areas overlap?
Sources
- Web Content Accessibility Guidelines 2.2
- Google general structured data guidelines
- Introduction to Lighthouse
- OpenAI publishers and developers FAQ
Version history
- 2026-07-15: Phase 8 maintenance, freshness, discovery, and ownership policy published.