>>Sounds good to me.  The whole notion of "aliases" seems silly to me,
>>really... they're just addresses on an interface.  I think the only
>>reason we have "alias" is because ifconfig and its ioctls were poorly
>>designed in the first place :-)

	bsdi4 changed ifconfig(8) to some degree to achieve this.  i have
	trouble remembering the exact syntax and/or the behavior.

>Um, well, there is the question of what IP address gets stuffed into 
>a packet without one (i.e. most applications don't bother to set one, 
>so you get one from the interface that the routing table decides to 
>send the packet out), and that marks the difference between the 
>interface's "address" and an "alias".

	4.4BSD IPv4 implementation basically assumes that there's only
	one address on a interface.  the code that manipulates rt->rt_ifa
	is one example, and the whole notion of "alias" is another example.
	not sure if it is a good thing or a bad thing.  if we want to change
	this for IPv4 codepath, we need a whole lot of changes.  IPv6 removed
	the assumption, and we have a lot of fun/discussion with source address
	selection issues.

itojun