Koogra was barfing on a bunch of Excel files I needed to
read that contained a mix of Chinese, Korean, and
The problem was Koogra was decoding strings 1 or 2
bytes at a time using Format.Convert, which didn't
handle 16-bit characters -- compressed or
uncompressed -- properly, and didn't handle surrogate
pairs (32-bit) characters at all.
The new behaviour is to read data for strings into a byte
buffer, then decode using System.Text.UnicodeEncoding,
like God intended.