> it comes from the fact that Office is ancient and very, very complex with a huge number of features.

IIRC one of the many unfortunate decisions made by MS with OOXML (whether intentionally, or not, or both) is to codify a lot of display and formatting quirks directly in the schema with very little explanation or docs. Instead of making it s different namespace or layer.

So, to implement OOXML, you also needed to reverse engineer, say, behavior of Word97 etc.