> This is I think correct. Initially I have no specific routing information
> to ax0 apart from the device configuration above, but this is sufficient
> to route access to all hosts on the 44.133.228/26 network.
While I am not up on "the networking rules" this I would consider to be a
Bad Thing. I'd say all configurations of Unix have in them an ifconfig
statement and then a route add statment straight after that for the network
directly connected on that interface. Now Linux may be different, I will
try to find out (probably in the next message in your mail-list).
I'm surprised that you can route with no routes, if so this has definitely
changed since I last looked at the Linux code, as if you brought interfaces
up but had not route commands, all hosts, including yourself were
unreachable.
> However another host on the same network, ea4rct, using rspf is
> broadcasting a route to THIS network on the same frequency. Here I am
> unclear whether this is correct, tolerated or not. The end result is my
> rspf daemon notes this information and adds a dynamic route to
> 44.133.228.0/26 through ea4rct.
>
> Having done so, any paquets, including broadcast paquetes are sent
> directly to EA4RCT, and not to the specified host on the network,
> without routing.
>
> Should not the "metric" of the interface's configuration
> take priority over a dynamic "learnt" metric, which has to pass through
> the same interface, or is the problem that ea4rct should not be
> broadcasting routes TO this network ON THIS NETWORK?
No, rspfd doesn't do that. It basically only looks at the routing table and
because when it was written there wasn't proper support for metrics, it
couldn't judge on metrics.
The rule it follows is "do not replace any non-dynamic routes already in the
routing table". This will change soon, now that metrics are in there, so
that it looks at the metric (which is what it should of done in the first
place). This interface metric will have to wait for more information.
- Craig vk2xlz