From Newsgroup: comp.misc
HererCOs a fun scenario to think about.
IrCOm setting up a replacement home router for a client. The existing
one is installed on their LAN, which I can get to remotely via
OpenVPN; the new one is in my living-room, connected to my laptop via
a dedicated Ethernet cable, on a different subnet from my house LAN
(which the laptop remains connected to).
From my main office machine, I can access either one via an SSH
tunnel; one to a machine at the clientrCOs place, the other to my
laptop. So I point my browser at rCL
http://127.0.0.1:-2suitable-port-nr-+rCY
to login and access the admin functions.
But when I try running tunnels to both at once (using different local
ports, of course), a peculiar thing happens: the login sessions tend
to expire within a few seconds, forcing me to keep logging in again
all the time. I get the feeling each router is seeing some of the
otherrCOs traffic (automatically-generated session ticklers, perhaps?)
and getting confused by it. Or maybe the ticklers arenrCOt getting
through at all, because they are omitting the port number ...
So I tried another approach, involving setting up routing entries to
get to the respective local subnets, with a little help from socat to
get across the last hop to the clientrCOs installed router. This way,
each router sees itself being accessed via a different address.
Now a different problem emerges: the router refuses connections,
instead showing an error page indicating it has detected a rCLDNS
rebinding attackrCY. I suspect this is because I am using one private
address range to access it, while it is configured to work in a
different private address range; if it saw itself accessed via a
public address, it probably wouldnrCOt worry.
Oh well. I think there one or two more combinations I can try ...
--- Synchronet 3.22a-Linux NewsLink 1.2