[OLSR-users] LQ, HNA and NAT

[Andreas Tønnesen]

Stefan,

You are correct in your assumption. When I did my master we suggested a
solution suggesting IP-in-IP tunnels to the gateways to be able to control
gateway usage from the actual host. In a OLSR routed network one has no
way to control what HNA gateway is used, and this can lead to many
problems.
(http://www.olsr.org/docs/report_html/node188.html)

The problem with changing HNA routes slowly as you suggest, is that HNA
routes follow regular routing, that means we cannot diffrentiate a route
to host A for a route via host A. Routes to gateways must be recalculated
just like routes to hosts.

It would also be interesting to add extentions to the HNA messages where
gateways declared stuff like bandwidth and current load of the uplink.

- Andreas

> Hello,
>
> I have one other question related to LQ routing: if LQ routing is also
> used for an internet-gateway HNA, mostly NAT is used. Now iirc for tcp a
> NAT usually creates a soft-state for the connection with the first SYN
> coming from inside (the NATed network). Now if the HNA route happens to
> change during a connection because the link quality changes, and another
> HNA with another NAT is used, all further outgoing packets may not get
> through to destination as there is no connection entry in the other NAT.
> Is that correct or am I missing something here?
>
> If that is true, it would maybe make sense to change the HNA routes very
> slowly in order to not break too much connections.
>
> In practice here in Berlins olsr-net I think this has not been tested
> because the version of the fff that most people use here still does
> calculate HNA on hop count.
>
> ...but I might be completely wrong here as well.
> Stefan
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users
>


---------
Andreas Tønnesen
http://www.olsr.org