>I agree that the mapped addresses are problematic/ugly (and I'd be
>happy to weigh in on that side of the argument on an appropriate
>working group mailing list..)..  but there's a higher-level issue
>here..  "Standard is better than better".
>
>Unless there's a *really* good reason for it, I think our stack should
>behave by default as specified.

	I'm more worried about having this rfc1933 mapped addr hack forever
	(or long time), in addition to security issue.  I believe this is
	not a good way to make applications dual-stack ready, since the spec
	is not clear enough about gory details like behavior of setsockopt,
	inpcb lookup ordering, and such things.  The hack looks fine at a
	first glance, but as you dig it deeper it has big trouble inside.

	If many of *bsd do not support it in default binary, people will use
	AF_INET/INET6 more properly when they port applications.

	I'll of course try to convince ipngwg folks in (possible) 1933bis
	discussion, in parallel.

itojun