The class IPTCInfo now has an inp_charset and an out_charset attribute - the
first is the read image's charset (defaults to the system default charset),
the second is the charset the writer will use (defaults to inp_charset).
Reader will find the charset included in IPTC data (if any, defaults to the
system's default charset), and use it to read to unicode strings. Writer will
write using IPTCinfo.out_charset (if it is not set, will not write charset
IPTC record).
With this, it is possible to read and write i18n strings correctly.
I haven't tested this functionality thoroughly, and that little test was only
on my WinXP box only, with the only other IPTC reader: IrfanView.