[ ] accept/accept4 seems to not be intercepted!
[ ] Clarify deny/fake for bind. It is only ipv4/ipv6?
[ ] Document force_poll_timeout
[ ] Split binding on connect from the one used for listen.
[ ] If we have more bind calls, we have a problem. Probably we have to use a table like this:
	#oper	family	type	address/port	address/port
	bind	ipv4	dgram	0.0.0.0/80	127.0.0.2/8080
	bind	ipv6	stream	::/80		::1/8080
	bind	unix	stream	/tmp/sock	/tmp/sock1
	alter	*	stream...
[ ] Test UDP sendto and sendmsg!
[ ] Probably we have to hijack also dup/dup2/etc.
[ ] Bandwidth per total or per socket. Better, per ip/protocol
[ ] Corrupt data
[ ] Add possibility to change destination of packets.
[ ] MTU
[ ] Add posibility to change on-the-fly the parameters by using a AF_UNIX socket
[ ] bwtop: http://superuser.com/questions/32932/is-there-a-way-to-find-out-what-application-using-most-of-bandwidth-in-linux
[ ] http://superuser.com/questions/106781/how-do-you-measure-the-bandwidth-used-by-a-given-process-after-it-has-completed
[ ] Bandwidth limitting also for download
[ ] Dump statistics (bytes, syscalls etc.) in a file specified by a env var.
[ ] Need to study how to set only the class for IPv6. Also check Linux kernel
	to see how prio is set for IPv6 class.
[ ] Nice: Force a protocol on a connection! So, user tries TCP but gets UDP!
[ ] Controlling timeout (SO_RCVTIMEO, SO_SNDTIMEO)
[ ] Enforce connect timeout
[ ] Change data on the fly!!!
[ ] Allow saving stream data in a file.
[ ] sendmmsg/recvmmsg support
[ ] There are stupid applications (skype) that run poll with timeout 0! Fix it!
[ ] Remote network: transport all sockets operations over another socket to a
	remote host.
[ ]