> Bump.         
> patch before it can be applied?
>
> Thanks!

Hey, you said matz has an another idea. So, we need to hear it. isn't it?

Personally, my feelings are,
 - read_nonblock and try_read_nonblock have too small difference to
make separate method. then, instead, I'd prefet to add an keyword
argument to read_nonblock.
 - I can imagine IO::WaitReadable is annoying to use. but I'm not sure
how much affect it a performance. Did anyone mesure it?
 - We need doc update about "select + readpartial is unsafe" issue.
Can anyone volunteer it?

And, for clarify, I think nobody replay Eric pointed following issue.
Although I'm not sure it is important or not.
 - Yehuda's original patch don't accept_nonblock and connect_nonblock.
 - current write_nonblock return written byte. but it is unuseful.
current String is designed intentionally byte operation unfrinedly.