[CUWiN] assigning numbers to a community wireless mesh

[David Young]

On Mon, Dec 06, 2004 at 11:54:35AM -0600, Stephane Alnet wrote:
> Hi Dave,
> 
> Before I say anything more, let me say I think you've a great idea. 
> I'll just highlight two points.
> 
> >So there's no harm using
> >link-local addresses for the routing protocol packets.  There's also
> >no harm using link-local nexthops (the OSPFv3 RFC recommends it, IIRC).
> 
> The only drawback I see is that it makes troubleshooting a little bit 
> more difficult because it breaks traceroute(1). But I'd agree that the 
> simplification in management of IP addresses is probably worth the 
> trouble.

Hmm.  Breaking traceroute(8) is pretty serious, since it is one of our
key diagnostic tools.  I see that both traceroute(1) and traceroute6(8)
have options -s src_addr.  If we set the source address to the IP address
on the ethernet, then the ICMP TIME_EXCEEDED messages will be returned.
However, the traceroute will show a bunch of link-local addresses.  Yuck.

I am beginning to like the idea of assigning IP addresses in
<site-prefix>:ffff/64 to the wireless interface.

> >The address
> >on the ethernet, however, had better be unique.  I think we should use
> >the ethernet's address to identify the router.
> 
> That's fine too, for the explicit purpose of selecting a routing ID.
> 
> Where I have a problem is that with this scheme you don't have any 
> _management_ address for the routers. The Ethernet address does not 
> qualify because if the interface goes down (for example because the 
> switch/hub/PC connected to it is shut down for any reason), then you 
> have no way to communicate with the device (can't talk to a down 
> interface), even though you can route to it.

I agree that if the Ethernet is administratively down, then we cannot
reach the router.  But if the switch/hub/PC is down, that's a "no
carrier" condition.  I am pretty sure that we can still reach the router
at its ethernet's IP address, even when the ethernet is in a "no carrier"
condition, provided we're talking to the router through its wireless NIC.

> You may want to think about using a loopback interface (or 
> sub-interface, depending on the capabilities of the OS) and assign it a 
> unique address (you only need one per device, no subnets, and they 
> don't have to aggregate nicely since HSLS is supposed to take care of 
> routing them). That's a pretty common setup. (IOS for example will use 
> a loopback interface's IP as the OSPF ID if one is configured, for that 
> explicit purpose.)

I think that someday, aggregation will be important, mainly to conserve
memory.  But I think that before that, multihoming will be important.
IPv6 routers and hosts will need to be able to change homes without
dropping TCP connections.  When they can do that, it becomes possible
to re-number the network so that it aggregates nicely.

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933