> Hoorrah!  Is some sanity checking in route(8) required/worth considering?

This is a "enough rope to shoot yourself in the foot" feature.  the
kernel verifies that both -ifp and  -ifa addresses point at real
interfaces; I'm not sure what else is necessary.

> Likewise, how would this affect routed/gated?

On first glance, it doesn't appear as if they would be affected at all.

routed:
	routed doesn't set RTA_IFP or RTA_IFA in the "route add"
	messages it sends, so existing behavior would be preserved.

gated:
	gated appears to set RTA_IFP in some of the multicast routing
	code for an RTM_CHANGE message (in krt_resolve_cache in
	krt_ipmulti_rtsock.c); my change is only to  how RTA_IFP/RTA_IFA
	are handled in an RTM_ADD message.

	like routed, gated doesn't set RTA_IFP or RTA_IFA in the
	unicast routing code.

Naturally, some additional work in the kernel and in gated would allow
for true unnumbered point-to-point interfaces.

					- Bill