From: Kent Sibilev <ksruby@...>
Date: 2005-05-25T01:39:36+09:00
Subject: Re: UDPSocket#recvfrom is slow

You can try Socket.do_not_reverse_lookup = true

Kent.

On 5/24/05, Niklas Frykholm <niklas@grin.se> wrote:
> The #recvfrom call in this code:
> 
>     socket = UDPSocket.new
>     socket.bind(0, 12345)
>     loop {puts socket.recvfrom(1000)[0]}
> 
> Runs extremely slow on my ruby (ruby 1.8.2 (2004-07-29) [i386-cygwin]),
> on the order of 2 seconds per call. I see similar problems have been
> mentioned before:
> 
>     http://www.ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-talk/136959
>     http://www.ruby-talk.org/cgi-bin/scat.rb/ruby/ruby-talk/143448
> 
> but no solution or diagnostic was given in the replies.
> 
> If #recvfrom is changed to #recv, the code runs at the expected (fast)
> speed, so I believe that the slow operation is the conversion of the IP
> address returned by C's recvfrom() to a network name. Apparently, on a
> (badly configured???) Windows network, this can be a really slow operation.
> 
> Is there a way of disabling this name lookup in the call to #recvfrom?
> If not, maybe one could be added in a future version of ruby.
> 
> // Niklas
> 
>