Re (tilly) 1: Socket Differences between Win32/Darwin and *nix.

I haven't done very much socket programming, but I suspect that an issue is that on Unix, a bound socket is not immediately released when you quit. (I have certainly run into behaviour like this before.) Therefore there is a (possibly considerable) delay after successfully binding and then quitting before anything else can successfully bind to that port again.

Therefore I would suggest organizing your server in a more traditional way. Rather than returning out of your accept loop, plan to get into an accept loop and then stay there servicing more requests until your program is done.

Comment on Re (tilly) 1: Socket Differences between Win32/Darwin and *nix.

Replies are listed 'Best First'.
Re: Re (tilly) 1: Socket Differences between Win32/Darwin and *nix. by clintp (Curate) on Jan 10, 2002 at 19:30 UTC
Excellent articles on this delay after closing a socket on a bound port are linked from here. OP: Read the FAQ articles (some are linked in the follow-ups) and if you decide that you want to bind on the port again immediately, use setsockopt and monkey with the SO_REUSEADDR option.	[reply]
(tye)Re: Socket Differences between Win32/Darwin and *nix. by tye (Sage) on Jan 10, 2002 at 20:34 UTC
Or simply add `Reuse=>1` to your `IO::Socket::INET->new` call. Update: ...which was already in the original code. Which makes me wonder how this particular error could happen... - tye (but my friends call me "Tye")	[reply] [d/l] [select]