Todo · BHG-025
Prototype responsive booklet reader
Goal
Create a repeatable conversion engine that turns native Publisher filtered HTML into semantic, responsive reading editions without losing booklet text, images, captions or page references.
Source
- Booklet implementation plan — stages and validation details.
- Booklet artefact inventory — 38 recovered working sources, 626 native-HTML pages.
- Native Publisher HTML and PDFs under
D:\bhg-booklet-recovery-20261019\.
In scope
- Consume the already captured and verified Publisher page-order manifest (626 pages, 0 missing). Validate the mapping against input files and preserve stable citation anchors; filename sorting is not authoritative.
- Parse Office/VML HTML into ordered text/figure/caption blocks. Preserve source refs, stable page anchors and source file metadata.
- A clean intermediate representation, reusable renderer and override mechanism for exceptions.
- Pilot with Johnny Green, The Burra Cemetery Walk and The Burra Story.
- Desktop/mobile rendering and source-text/image QA, recorded in a report.
Out of scope
- Public site Preview/Real.
- Bulk processing all other booklets.
- Uploading PDF/Publisher binaries or publishing content to House Sites.
Acceptance criteria
- Three usable responsive pilots preserve all extractable source HTML text and 178 source image uses, with zero missing assets. 42 image-based pages and caption/alt associations remain explicitly flagged for editorial work under BHG-027, not silently described as complete searchable text.
- Page order taken from saved Publisher PageID manifest (138/138 source pages in these pilots).
- Semantic article/page markup, true table markup, focusable navigation and no horizontal scrolling at 390px mobile width.
- Rebuild script, validator, structured JSON output, per-page editorial override mechanism and documented QA report exist.
- Source-only reading, Publisher SHA-256 checks, output solely under a separate Dev working area.
Stop and ask
If reliable page order or content fidelity cannot be established, report where automated handling stops and what manual work is needed rather than publishing a lossy reader.
Completion notes
Completed 19 October 2026. Detailed pilot report. Generator and tests: bhgsite-working/scripts/booklets/.
- Three readers generated locally: Johnny Green (4 pages), Burra Cemetery Walk (26), Burra Story (108).
- QA passed: 138 pages, 817 ordinary source text blocks, 15 table-cell paragraphs, 178 image uses, one reconstructed data table, 0 missing assets; verified 390px mobile no-overflow on each.
- Recovered the five-row mortality table on Cemetery Walk page 9, previously outside Publisher's normal text shapes.
- PDF/Publisher originals unchanged, no working-site or Preview/Real deployment in this ticket.
- BHG-027 must address predominantly image-based pages (42 in the pilot), source caption/figure associations, multi-column layouts and reviewed image descriptions. Do not market the pilot as complete editorial transcription.