[Olsr-dev] Patch Orgy Syncup

[Juliusz Chroboczek]

> the real problem scope is not so much the matter of protocol choice
> but rather fixing the problem of "my default gw does NAT" ...

Don't do NAT, then?

> clearly the source-routing protocols have an advantage here, however
> as mentioned before there is tunelling which gives the ingress some
> control.

Yep.

The alternative would be to have all border routers (gateways) tunnel
to a single NAT box.  The advantage being that you get optimal routing
within the mesh cloud, the downside being that you get a single
bottleneck and point of failure at the NAT box.

> introducing an auto-tunneling infrastructure would also be beneficiary
> for the phased approach of V6. (olsr speakers shall automatically create
> V6 tunnels in order to connect islands) - i really dilike the prevailing
> all-V6 or nothing approach ...

I agree, although my preference is to do the opposite -- route IPv6
natively, tunnel IPv4 in IPv6.  There's a lot of things that are much
easier in IPv6 than in IPv4.  The biggest win is the availability of
link-local addresses, which can greatly simplify routing.

(Advertisement: http://www.pps.jussieu.fr/~jch/software/ahcpd/ , the
all-dancing, all-singing autoconf daemon for mesh networks.  Yes, it
does OLSR too.)

                                        Juliusz