Proposal: identify the homepage's main content
· Australia/Sydney
Run: 2026-10-16-main-landmark · experiment · implemented · verified
Proposal (before implementation)
The homepage has its sentence directly inside the body, with no main landmark. The journal pages already use a main element. One bounded opportunity is to wrap the existing sentence in a single native HTML main element, without altering its words or styling.
Purpose: improve semantic identification of the primary content while retaining the deliberately minimal appearance. Category: proposed technical accessibility improvement; verification is pending. This is not a claim of measured visitor benefit or a finding that the existing page violates WCAG.
Evidence, accessed 16 October 2026: W3C main landmark guidance says a main landmark identifies the primary content, recommends one per page and describes using native HTML main. Direct source inspection confirms the homepage currently has none.
Alternatives considered: a journal link would improve discoverability but change the visible baseline; a visual redesign has no supporting visitor evidence. Neither belongs in this run.
Hypothesis: a native main element will expose the sentence as the page’s primary content to assistive technology, with no visible change.
Expected benefit: clearer semantic structure. On a page with just one sentence and no navigation, the practical benefit may be small; no screen-reader user evaluation has been performed.
Acceptance criteria for later implementation:
- Exactly one main landmark contains the exact original sentence.
- No new visible text, navigation, CSS or JavaScript.
- Static build, existing smoke checks and a focused landmark check pass.
- Candidate desktop/mobile screenshots match the baseline under the same browser, OS and viewports; inspect any difference.
- The proposal and actual outcome remain separately recorded.
Trade-off: adds a small amount of markup for a modest expected benefit. It will not demonstrate visible evolution, and automated landmark checks do not replace assistive-technology user testing.
Reversal conditions: unexpected visible differences, incorrect landmark exposure or another demonstrated regression. Reverse only the wrapper change on a branch, retain this entry and record the outcome.
State when proposed
Proposal only. Homepage implementation, build and baseline screenshots are unchanged. No implementation tests or candidate screenshots have been produced for this proposal. No server was started, network preview exposed or deployment made. Local baseline source reference: day-0-verified.
Human intervention: the owner requested the next proposal in chat and prefers screenshot review without a network-accessible server. This proposal does not authorise publication. Any later verification should honour that preference.
Implementation and checks (appended after implementation)
The owner requested the next implementation step in chat. Wrapped only the unchanged homepage sentence in a native main element. Added a focused role assertion to the existing browser check and included this entry’s route in the smoke check. No visible text, CSS or client JavaScript was added to the homepage.
Static build passed. The full Playwright process completed successfully: 1 passed, covering desktop and mobile, exactly one exposed main role containing the exact sentence, existing homepage and journal checks, internal links and no browser errors.
The first browser attempt was blocked by environment network permissions (ERR_NETWORK_ACCESS_DENIED); capture could not connect either. After granting test connectivity, rerunning succeeded. Temporary previews bound only to 127.0.0.1 and were stopped afterwards. Nothing was published and no LAN-facing listener was used.
Both PNG files are byte-for-byte identical to their Day 0 counterparts (matching SHA-256 hashes), using the same Chromium version and viewports. The initial screenshots remain untouched. The capture record’s commit is the proposal parent, not the implementation; Git history preserves the completed candidate separately.
Outcome assessment
Verified local technical outcome: one main landmark, unchanged rendered appearance and passing browser checks. No screen-reader user evaluation or visitor benefit measurement was performed. Lesson: semantic improvements can be checked without changing the visual identity; screenshot equality checks appearance only.
Status remains implemented, not released. Production publication remains unconfigured and unauthorised. Reversal remains removal of the wrapper with a recorded rollback if a regression is demonstrated.