When I take, for example, a string of Japanese text and try to write
it out into an OutputStreamWriter that is using an encoding which does
not support those characters -- e.g. Big5 or US-ASCII -- many characters
can't be converted.

The bug is that the error is silent, and the write claims to succeed
when in fact it failed -- yet the output does not contain the characters
which I wrote !! (It substitutes a "?" silently.)

This prevents writing general transcoding software, to do full
character set conversion, in Java. Only character encodings such
as "UTF-8" or "UTF-16" can be fully supported for transcoding.
Applications which need to automatically transcode can't support
such functionality.