>Is there a way to specify both tcp v4 and tcp v6 at once for when you
>want both to be served?

	RFC2553 says that wildcard-bind to IPv6 socket will accept IPv4
	traffic, as if it was from "::ffff:10.1.1.1".

	If you run only one "tcp6" server on port X, connections to IPv4 port
	X will be routed to that server.

	There are several issues remain:
	- for IPv6 UDP, wildcard-bind behavior is not implemented yet in
	  the kernel.
	- ftpd cannot handle IPv4 connetions from IPv6 socket.  You need
	  two lines on inetd.conf (for "tcp4" and "tcp6").
	- we may need to have a way to disable the special behavior of
	  IPv6 wildcard bind (setsockopt?).  It is not on the document.


	Personally, I'm very against about the way wildcard-bind
	IPv6 socket is spec'ed.
	It seems to ease your life first.
	However, actually it makes your life much harder.
	- name resolution is way more complex: see RFC2553.  many pages are
	  filled with special behavior of name resolver.
	- If your application creates sockaddr from some outside data (like
	  ftp: it needs to construct sockaddr from PASV/PORT parameter)
	  you'll need to support 3 kind of addresses:
		AF_INET socket, normal IPv4 address
		AF_INET6 socket, normal IPv6 address
		AF_INET6 socket, mapped IPv4 address (::ffff:10.1.1.1)
	- it is a special case, which is not future adaptable.  What happens
	  if you have IPv7 in the future?  how will the wildcard bind behave?
	- port number issue and bind order issue is never described.
	  What happens if you make a IPv4 specific bind for port X, then
	  IPv6 wildcard bind for port X?  There are so many possibilities
	  and there's no document.  No way.

itojun