link-local needed w/static IP and gateway?

[Charles Sprickman]

Hello,

I've been trying to wrap my head around the differences between address 
resolution in IPv6 and IPv4 and I'm a bit confused by a real-world issue 
I'm seeing in a colo facility where we have dual-stack connectivity.

Basically, what I see is summed up in this recent post:

http://freebsd.1045724.n5.nabble.com/Proper-way-to-setup-IPv6-gateway-on-running-node-without-reboot-td4313847.html

If I manually configure a static IPv6 IP and then set a default route for 
a router on the same subnet, ie:

ifconfig em0 inet6 2001:xxx:xxxx::2/48
route add -inet6 default 2001:xxx:xxxx::1

I have no issues pinging other hosts on the subnet (which also have static 
IPs and manually configured gateways), but I find that address resolution 
for the router is spotty at best.  If I start and maintain a ping from the 
host to the router, the first few packets are lost, then traffic flows. 
If I'm pinging from an outside host, once I stop the ping from the host to 
the gateway, the external ping fails shortly thereafter.

I can also get traffic to flow to the gateway briefly by running "rtsol 
em0", but after a few minutes it stops.

Now following the steps in the thread linked above works, and what that 
basically has you do is enable link-local addresses, down/up the 
interface, and then all is well.

Can anyone help me understand what the relationship is between address 
resolution for the router and link-local?  Why is this required?  Why can 
I ping other hosts on the subnet without enabling link-local?  I 
understand link-local is needed for *automatic* router discovery, but in 
my case I'm explicity setting a default route.

I'm having a hard time finding good docs on this, most tutorials seem to 
center around a tunneling setup or simple autoconfigured LAN stuff - no 
one's really addressing typical colo/datacenter configs.  I've got my 
workaround, but I'd like to understand what's going on.

Thanks,

Charles