First, we don't care about/want malformed characters replacement that happens with new String(byte[], Charset) / StringCoding because it's a great way to silently fail and never realize there's encoding issues.

Then, assuming above statement, Netty's ByteBuf.toString(Charset) generates lots of garbage:

• It can only decode one single ByteBuf so if we have several of them (chunks), we have to wrap them into CompositeByteBuf.
• As it wants to enable replacement, when dealing multiple underlying NIO ByteBuffer, it has to merge/copy them into a single ByteBuffer (which is/can be pooled).

In the case of US-ASCII and UTF-8, we can avoid this copy and use the CorderResult status to detect char split between several buffers and analyze the first rejected byte to determine how many bytes we need to complete the char.