[OLSR-users] LQ, HNA and NAT

[aaron]

I would like to elaborate a bit on OpenVPN tunnels to a single NAT 
point:

Why do we have it:
   no better solution so far

Why is it bad:

1) SPOF
2) single point of control - (politically bad :)

Let me elaborate on this a bit, with our setup:

    A --> OpenVPNgw1  -- tunnel over cable --> OpenVPN server + OLSR   
<--- OLSR client C
                                                  ^
    B --> OpenVPNgw2  --- tunnel over cable --> - |

OpenVPNgw1 will get default routes via OpenVPN server (runing OLSR)
Therefore It can only have a static host route to the OpenVPN server 
via its only next hop (which used to be its default gw)
--> When the OpenVPN server is down (like it was today) , you cannot 
reach the OpenVPNgw anymore from somewhere else


in numbers (intentionally changed a bit):

OpenVPNgw1: 62.116.0.777

OpenVPN server: 62.116.38.66

OpenVPNgw2: 1.2.3.4


routing tbl in OpenVPNgw1:

dest             gw
62.116.38.66     62.116.0.778  (<- this used to be the default router)
0.0.0.0          62.116.38.66  (<- this came from OLSR)

So you can see, if 62.116.38.66 is down for some reason, the OpenVPNgw 
is also "down".
Bad.

I thought about this problem a bit and - yes i do believe we need NAT 
as long as we dont have IPv6 everywhere,
but on the other hand it would definitely make sense to have some kind 
of fail-over behavior between inet gws.
Something like multicasting NAT entries is also not nice on the other 
hand.

Question: Can anybody see a problem with giving official IPs to "last" 
hops only? Then the "inbetween nodes" can have internal IPs, everybody 
would "talk" to HNA nets/hosts only and NAT can be abandoned.
How would the routing tables grow? Every HNA entry in every inbetween 
node as extra info? Doubling the routing entries?

ciao,
a.