[Olsr-dev] IPv6 Unique Local Addresses

Alexander Morlang (spam-protected)
Sat Nov 29 01:01:51 CET 2008

Hash: SHA1

Alina Friedrichsen schrieb:
> Hi John!
>>> In OLSRv1 (that is what we are using today) this would work, but
>>> ipv6
>> itself
>>> is a problem with OLSRv1 because the address fields are getting
>>> HUGE.
> 4 Bytes <---> 16 Bytes?
> I think using a FPU on devises, that don't have a FPU was a much
> bigger problem. Jumping into the kernel space for every calculation
> costs much more then 3 instructions more for every IP address.
>>> In OLSRv2 (which can use address compression for address blocks)
>>> even
>> with
>>> "autoconfig" it would be better to use a real subnet so that you
>> compress the
>>> blocks (common header/suffix-compression).
>>> But the IPv6 autoconfig might be a good step to get a "permanent"
>> adderss in
>>> an OLSRv2 IPv6 network. :)
> The plan is to use the unique local address only for bootstrapping. I
> think to write a autoconf algorithm for it is a little bit oversized.
> A 64 bit random number does the same, with less problems. If now the
> node is in the mesh it searches for the nearest Internet gateway. If
> found, it takes his/her 64 bit random number in the unique local
> prefix and put it after the global prefix which the Internet gateway
> routes into the mesh. So the node has now a real public IP address
> with which is now reachable from the Internet without NAT, a
> centralized or other suck.

i think, interface configuration is not the job of a routing daemon, if
you want to do stateless ipv6 autoconfiguration in a mesh, you can use
an protocol like ahcp (http://www.pps.jussieu.fr/~jch/software/ahcp/),
which also gives you a global routeble prefix before starting the olsr
daemon (if you want it).

> With the unique local address of the node address / routing table
> compression should work not so good but with the public global
> address in the Internet it should work fine.
>> Something that should be considered is to use link-local addresses
>> for the olsr packets. The ipv6 addresses that are exchanged should
>> still be global or the unique local rfc4193 style. What I'm talking
>> about is the address of the packet itself. There are two advantages
>> that I can see.
> I don't know what from-addresses packages to multicast addresses have
> or should have. :/
>> 1 - It is closer to how other routing daemons work.
> Which daemon?
>> 2 - You automatically have the interface info because it is part of
>> the link-local address.
> Do you mean the MAC address?
> Regards Alina

Version: GnuPG v1.4.8 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the Olsr-dev mailing list