[Cu-wireless] phase iii network: v4/v6 addressing & Internet connectivity

[Stephane Alnet]

Dave,

Regarding the justification of using IPv6...

>   2 Routable IPv4 numbers are in short supply. It is not clear
>     whether CWNs can get hold of them, and it appears that they will
>     cost a *lot*. [...]
>   3 IPv6 is unquestionably the future of the Internet.People are 
> already
>     using v6 a little bit[ ...]

[IPv4 addresses are basically free (granted, there's the ARIN fees), so 
you can't really use that as an argument. But that's beyond the point.]

I think that:
- being an experimental network using free software, we don't really 
need to justify our choices in terms of technology beyond the technical 
arguments;
- but even if we did have to, there's probably a bunch of stuff already 
written by various marketing departments (like [1]) that should do just 
fine if we have to convince anybody of the virtues of IPv6.

[1] http://www.cisco.com/warp/public/732/Tech/ipv6/docs/sod.pdf


Regarding the routing part of the proposal, I need a drawing to better 
understand it, but it sounded like you're more-or-less trying to do 
something similar to MPLS. With additional complexities like 
IPv6-over-IPv6 and dynamic associations ("supplicant", etc.). If I 
understand correctly, you are trying to create classes of supplicants, 
the equivalence relationship being "routes traffic outbound through 
gateway X" [whether you use this to create tunnels or tag traffic 
à-la-MPLS is really an implementation artifact in my opinion, not a 
design limitation]. I'm a bit confused at how the traffic internal to 
the network gets handled, however (that's the traffic that has no 
outbound gateway) -- obviously it would get routed "normally" but are 
there any border conditions that need to be taken care of? I guessed 
that IPv4 traffic gets nat'ed and IPv6 traffic is tunneled, but that 
may not have been the intent either. Just checkin'

I tend to believe that nd/nsd-kind-of things would be better served by 
ideas you already put forward (like using multicast, or maybe 
well-known, network-limited addresses, now that we are talking about 
IPv6?), independently from the underlying network architecture(s). 
Nothing prevents us from having (some) routers or (better, given the 
architecture proposed) the gateways acting as servers; of course the 
case of the "no outbound gateway" class of traffic(/nodes?) requires 
special attention. I also think that providing the (larger) community 
with an IP-based, self-bootstrapable "name" service system (aka 
directory) would be a great realisation, whether or not you solve the 
routing (& other) issues -- that's another reason I'd see to disconnect 
this from the routing protocol development.

Just my non-expert opinion, YMMV. :)
S.

--
Mayotte - http://stephane.alnet.free.fr/