Test text:
<!-- A tolerance of ±4.5 cycles a second is satisfactory. -->
When I paste text into disambiguation.xml or grammar.xml, I get a Java error:
MalformedByteSequenceException: Invalid byte 1 of 1-byte UTF-8 sequence.
If I remove the plus/minus sign, LT starts.
As best I can tell, ± is a valid UTF-8 character (Unicode Character 'PLUS-MINUS SIGN' (U+00B1)). So, why does LT/Java complain?