From: Gary Wright <gwtmp01@...>
Date: 2009-07-15T04:45:38+09:00
Subject: Re: Process.fork weirdness


On Jul 14, 2009, at 12:23 PM, Joel VanderWerf wrote:
> Ah, that makes sense. It would be only the libc and ruby buffers (in  
> user space) that would get copied in the fork, not network stack  
> buffers in kernel space.

That is correct. IO data buffered in the kernel is not duplicated by a  
fork.

> But wait... Net:HTTP uses BufferedIO#writeline on sockets. That  
> class doesn't itself buffer writes (only reads), but it is defined  
> in terms of IO#write, which is buffered, isn't it?

Ruby marks IO objects associated with sockets as 'synchronized' (i.e.  
unbuffered).  Net::HTTP wraps the underlying ruby socket with  
BufferedIO but my quick inspection of the code seems to indicate that  
BufferedIO doesn't introduce a write buffer, just some debugging  
scaffolding and a helper method for writing lines terminated with '\r 
\n'.