--- Joe Maimon <jmaimon at ttec.com> wrote:
Hello Rodney,
At first cut, I am trying to effect a seperation
between the interfaces
which need (overload)natting done and the ones that
dont. Exactly what
that will buy me in terms of nat problems,
performance or logical
correctness I am not quite certain yet.
As is currently, If it turn nat on for some
interfaces on the router, I
have to turn it on for all so that others dont see
rfc1918 that they
would not be expecting. Such is only proper.
Why nat? Well some customers like to link up a few
of their sites with
the cheapest CPE possible which supports the
simplest network possible.
A Linksys router is $40, and it runs NAT. I can't
really imagine that that's a serious cost barrier for
CPE.
Suggestion: have NAT happen on the CPE rather than on
your edge - it makes more logical sense, as that is
the actual point where the circuit becomes multiheaded
(i.e. many LAN devices, only one WAN path).
The solution you're looking at may actually work.
However, I can't imagine that you're not buying
yourself future trouble when you have to
migrate/scale/etc.
Also, running RIP with customers is another idea which
will encounter some serious scaling and security
issues pretty quickly.
-
9 messages in net.nether.puck.cisco-nsp[c-nsp] Re: Interfacing between VRF a...
