In article <6.0.0.20.2.20080917213249.03cbf078 / localhost>,
  Martin Duerst <duerst / it.aoyama.ac.jp> writes:

> Can you explain? Does it mean that on output, UTF-16BE
> is explicitly (e.g. via UTF-8) converted to UTF-16BE, if that's
> what's needed? If that's the case, why only on output?

ASCII incompatible encoding may be stateful as ISO-2022-JP.
An IO opened by "w:enc" needs an encoding converter to hold
the state.  A newline converter can be attached to that.

An IO opened by "r:enc" don't need an encoding converter.
-- 
Tanaka Akira