The close-all-issues run left several downstream refusals whose real blocker is the shared model, not any codec. Five fields, each named permanent by a scorecard this week:
- ContentRun carries no vertical alignment (superscript/subscript) and no text direction (RTL) -- the loss crosses to every format, not just one (rtf-codec and epub-codec's READMEs already document the identical gap from their own sides)
- no per-cell font field on the spreadsheet cell (xls-codec's writer; ooxml.js makes the identical choice for xlsx)
- no home for user-defined names (xls Lbl)
- no original compressed image bytes on LayoutImageAsset/ContentImageBlock (the read side normalises to decoded pixels) -- the sole blocker for JBIG2/JPX verbatim passthrough in pdf-codec's writer
Evidence
#971 and #975 scorecards; rtf-codec/epub-codec README fidelity notes; xls-codec README writer scope.
Affected packages: document-schema.js, xls-codec, ooxml.js, pdf-codec
The close-all-issues run left several downstream refusals whose real blocker is the shared model, not any codec. Five fields, each named permanent by a scorecard this week:
Evidence
Affected packages:
document-schema.js,xls-codec,ooxml.js,pdf-codec