All,
Andrew Jackson recently opened TIKA-1678. Tika tries to use Dublin Core
items from the xmp, and if that doesn't exist, it takes what it can find from
the "regular" metadata.
Andrew found that for ~200k out of 21million files, the UTF-16 is incorrectly
(? doubly?) encoded in the xmp :
\376\377\000B\000u\000l\000l\000z\000i\000p\000 \000P
Should we add a handler at the Tika level to deal with obvious BOM-marked
strings we're getting from the XMP, or should that be handled by PDFBox? We're
still using jempbox...will XMPBox handle these correctly?
Thank you!
Best,
Tim
