I've committed changes that add support an "encoding" attribute on
appenders that inherit from WriterAppender. The internal structure of
WriterAppender has been changed to bring it a lot closer to log4j.
With these changes encoding has been moved into CharsetEncoder and
CharsetDecoder which are patterned after the corresponding classes in
JDK 1.4.
The implementation currently uses the apr_xlate method in apr-util,
however apr-iconv which is used when there is no detected iconv (aka
Windows) seems to have problems. If I can't work through the issues, I
may be up and modify CharsetEncoder and CharsetDecoder to have Win32
specific implementation. At the present, the CVS HEAD is not usable on
Windows.
