I don't know. Socket.write() is hot code, it should be as fat free as possible. I'd rather solve this by making the docs more explicit.

@koichik?

@bnoordhuis

I'd rather solve this by making the docs more explicit.

Because most of functions such as a Buffer() accept encoding name by uppercase, I think that we should fix this problem.

But, it is not only 'utf8' but also 'ucs2', 'hex' and 'base64' that the number of characters and the number of bytes are different.
Therefore:

• Buffer.write() should always set _charsWritten.
• Socket._writeOut() should not care its encoding.

Or, as well as net_uv.js:

• Socket._writeOut() should create new Buffer rather than reusing it.

Buffer.write() always sets Buffer._charsWritten by #1633, Socket._writeOut() should not care its encoding.
@bnoordhuis - Can you review 55e6be1?

@AjClarke - This has just fixed in the another way. Thanks for the report.