> It is the way routing works. If you have static routes pointing to all
> of the IPs in a PM's pool, then *OF COURSE* it will bounce them back to
> the ether is the person is not connected. If you use RIP/Proxy-ARP the
> packets should never reach the PM in the first place.
>
> User diconnects, ARP entry goes away. Packet from outside world comes
> into your network, router can't find any machine claming to be that IP,
> dead end route.
>
Brian:
For some reason I'm having a hard time buying all this. I'll tell you
what I have at one of our locations and you tell me what I'm doing wrong.
We have a site with static routing only. The ethernet is fed by a
Cisco router which knows that this particular network is subnetted with
255.255.255.224 (30 useable addresses each). We have the ether0 ports of
3 PortMasters in subnet 1 at .34, .35 and .36. The assigned addresses
are set to .65, .97 and .129 respectively. The Cisco has static routes
for .64 to .34, for .96 to .35 and for .128 to .36. In this way, each
group of assignable addresses are a separate subnet and are directed to
the correct PortMaster. In this configuration, doing a traceroute to an
unused assignable address results in a routing loop between that
PortMaster and the Cisco router.
Perhaps I'm a little slow with this stuff, but that all seems to be a
very reasonable way of configuring it. Doing this let's us get maximum
utilization out of that class C and results in a couple extra subnets we
can deliver to users. In this situation of course, any packets for currently
unused assignable addresses DO get sent to the PortMaster.
Perhaps I made a mistake when I imagined a particular subnet for the
assignable addresses residing on a particular PortMaster. As part of
that possible error I'm using the ethernet port on the PortMaster as a
gateway to that subnet. If that's the case, then 255.255.255.224 as a
netmask won't work because with 30 addresses already in use in those
subnets, there isn't a free address to use as an alias on the Cisco's
ethernet port.
Now, thinking of how our PM2ER's are deployed, where we have a single
PortMaster with nothing connected to the ethernet port, we do have subnets
which reside entirely on the PortMaster and in the case of having the WAN
port unumbered, we are correctly using the ethernet port of the PortMaster as
the target for those subnets. Correct?
Hmmm, we still end up with the same problem. I can see how if I had
not subnetted the network where we have the 3 PM's together with the
Cisco, that I could have wasted some address space and configured it so
there would be no problem. But with the PM2ER's on the end of a WAN
connection, I don't see how I can have that situation.
Wow, I'm still confused. I just don't see how I can consistantly deploy
PortMasters without having problems with them trying to return unused
assignable addresses back to the default route unless I use a static
route to deliver them unceremoniously to the ethernet port's address.
Chuck