Claude Jones wrote:
I gather that it's simply a case of you moving from direct connection to
the Internet to a connection that's behind your Cisco router. I further
presume that the (nominal) 70.x.x.120 address belongs to the Cisco. In
that case you probably don't want to do what you propose.
Thanks for your thoughts, and yes, your surmising is correct
You might be able to make this work with proxy-arp from the router but
there are next-to-no tools to debug things at that level.
Your box should be set up on the internal LAN subnet in the normal way.
All of the interesting configuration should be done on the Cisco router
by setting up source and destination NAT so that internet traffic on
specific ports addressed to the Cisco are routed to your box and
responses are automatically routed back through the Cisco to their
destination.
I viscerally believe you're correct, here - else why is this the way this is
universally done, but I sure could use some better technically grounded
expertise in the whys and wherefores
Pretty much all IP routing concepts and diagnostic tools depend on the
idea that every two connecting points are on a common subnet or defined
as point-to-point. If you depend on proxy-arp and it doesn't work, all
of the diagnostics will say it's not supposed to work.
Trying to deal with this issue from inside the LAN rather than in the
router will most likely lead to frustration since whatever you do will
be fragile and probably break often if it works at all.
This is where I need better argumentation...if you can help, it would be
appreciated. Specific examples of why it's a bad idea, security problems that
could occur, other issues...unfortunately, this configuration has been handed
to me, it's not my idea, so I need to understand what's wrong with it and be
able to offer sound arguments for the more conventional approach, if there's
is a really sound technical reason for not doing it this way.
I'm also dealing with the fact that another Linux box, a mail server, has been
moved on to this new FIOS lan and configured using the hack that I cited in
my original post, and is working quite nicely - unfortunately, I don't
clearly understand how to implement that hack on Fedora, but, I'm getting
the "if Jack could do this with his Debian box, why can't you with your
Fedora?"....
I couldn't access that link you posted so perhaps it's not the best
place to get networking advice. What is the argument for not doing it
the well understood way? That is, if all of the addresses on the subnet
are on this interface or the ones you need can be split to a smaller
subnet, just route normally, or if you need a few addresses on the wrong
side, use static NAT? At least ask what diagnostic Jack will use to
determine where the problem is when his connection is down.
--
Les Mikesell
lesmikesell@xxxxxxxxx
