Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 24 Aug 2006 06:59:11 -0400
From:      Pat Lashley <patl+freebsd@volant.org>
To:        Ian Smith <smithi@nimnet.asn.au>
Cc:        freebsd-net@freebsd.org, Doug Barton <dougb@freebsd.org>, Fredrik Lindberg <fli+freebsd-net@shapeshifter.se>
Subject:   Re: Zeroconfig and Multicast DNS
Message-ID:  <4361C9398E9A195D2FB0B011@garrett.local>
In-Reply-To: <Pine.BSF.3.96.1060824160937.11983B-100000@gaia.nimnet.asn.au>
References:  <Pine.BSF.3.96.1060824160937.11983B-100000@gaia.nimnet.asn.au>

next in thread | previous in thread | raw e-mail | index | archive | help
> I've been watching this thread with great interest, having recently been
> introduced to the possibilities of OLSR (net/olsrd) for local (and
> beyond) P2P wi-mesh networks, and wondering if/how zeroconf fits in.
>
> Some refs: My discovery point, a great (online) book found from a review
> by Geoff Huston in the Internet Protocol Journal Vol 9 No 2, p44:
>
> Wireless Networking in the Developing World: http://wndw.net/
>  OLSR.ORG: http://www.olsr.org/
>  RFC: http://ietf.org/rfc/rfc3626.txt (basis, though olsrd extends this)
>
> Host addresses in such a MANET appear to require manual allocation so
> far, usually in RFC1918 ranges, but the notion of zeroconfig-joining
> such a network seems perhaps worthy of exploration?
>
> Am I way off base here, thinking some matchmaking might be useful?

This is all off the top of my head...

For a small enough mesh, with low enough latencies, I believe that you could 
just define the entire mesh as one link, sharing a single instance if the Link 
Local IP range.  Everything acting as a router/bridge would have to propagate 
the various LLA packets throughout the mesh but avoid sending them off-mesh. 
OLSP or other ad-hoc routing protocols should handle that setup as well as if 
every host had a static IP. (I don't remember the details of LLA, but I know 
that mDNS needs multicast support. So you would need an ad-hoc routing 
mechanism that supports multicast to get a full zeroconf mesh with DNS and 
service discovery.)


But the design would start to break down as the mesh grows large enough to 
either use a significant percentage of the LL address range or to make 
end-to-end latency significant.

I can think of a couple of potential approaches to designing a federation of 
smaller meshes; but they all have some pretty tricky issues to resolve.


I strongly suspect that it would be simpler to just build your mesh using IPv6 
only; and to provide 6-to-4 (NAT?) conversion at the interfaces between the 
mesh and the WAN. (The IPv6 address range is large enough that every interface 
already has a globally unique link-local address; so no need to negotiate, 
defend, or change the IP addresses as things move around.)


Also, I'm not familiar with OLSP; but I note that it apparently actively 
discovers nodes one and two hops away. It isn't clear to me how it handles 
routing to anything further than two hops. I also note that there are a 
mind-boggling number of ad-hoc routing protocols to choose from...



-Pat 



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4361C9398E9A195D2FB0B011>