>As an [original] author of the IPv6 Adv API, I have no problem with the
>sysctl.  But I just made a proposal to add a socket option to control
>this on a per socket basis.  This is because an application needs to
>be able to control this itself.

	Hmm, looks like:
	- have a sysctl to change system-wide behavior
	- have a setsockopt to change per-socket behavior
	- default value for sysctl is still under debate.  I vote for "off".

	matt, do you have any specific symbol in your mind for setsockopt?
	(kame/freebsd3 uses IPV6_BINDV6ONLY, which disables RFC1933 special
	behavior when set to 1)

itojun