>>	the following patch uses 0.0.0.0/8 as 24bit interface index value.
>>	on getsockopt(IP_MULTICAST_IF) it will return the value the user has
>>	set.  does it look right, or messy?
>	a question remains to me - how can we identify a system with this
>	feature, and a system without feature.  this is messy...

	another question, should we keep the old behavior 100% compatible
	by assigning a new setsockopt number and keeping the old one
	as IP_MULTICAST_OIF or whatever?

itojun