That's really really bad advice. You can get encoding errors like that on the java side. Yeah, the bytes in memory are all that matter and you're fine to interpret, but on the way in and out, you're playing with things.

It's the same thing like binmode. You're affecting the data as the IO occurs to get into our out of memory. See...

http://www.docjar.com/docs/api/java/nio/charset/UnmappableCharacterException.html

I've run into this exact problem using the XML feeds for perlmonks while working in java.