Maintaining a translated document corpus in OpenDocument

The format choice drives everything and was not really a choice. Matching a specific typeset layout means starting from a document that reproduces it rather than re-typesetting from markup, so the source is OpenDocument, edited and exported in LibreOffice.

ODT is a zip of XML, which means git sees a binary blob. No readable diff, no merge, and two people editing the same translation is a conflict no tool can resolve. Reviewing a translation pull request means opening both versions and comparing by eye, which is nothing like reviewing code.

The figures make it worse in a memorable way. The diagrams carry live text, so translating them means double-clicking into the image in Writer, or copying it into Draw and ungrouping to change box dimensions. Nobody would guess that, which is why the README documents it.

LaTeX or markdown with generated figures would be better in every way except the one that mattered, because it would not produce a document that looks like the original.

Translators are asked to work directly on the English ODT and link back to bitcoin.org/bitcoin.pdf, the right instructions, and unenforceable.