Lighttpd

While experimenting with netcat (the original 1.10 version, not the GNU version) I encountered a weird problem. Everytime I try to use netcat to simulate an HTTP client against an Lighttpd server, the process either hangs until timeout or dies unexpectedly. While hanging, the connection shows up in server-status in state 'read' with 0 bytes being read or written.

I tried netcat on three different platforms:

- On Cygwin 1.5 the netcat process mostly hangs, but sometimes it dies unexpectedly.

- On Fedora FC3 and CentOS 4.4 the netcat process always dies unexpectedly.

Never did I get any response from the Lighttpd server.

I made a script which illustrates this behaviour. First a hit is being done with curl and next with netcat simulating curl (sending the same headers etc). I've tried this on a few random sites and these are the results:

Now this could be a flaw in netcat ofcourse, but the question is: why would such a fault be triggered by Lighttpd and not by other webservers?

I have straces for both client and server, but I guess those can be easily reproduced with the attached scripts.