[E-Lang] Getting around NAT
Sat, 8 Dec 2001 10:33:23 -0400
I was wondering if there are any networking gurus here who might know
why the preferred methods for extending the IPv4 address space are
NAT for now, and IPv6 for later, as opposed to using the Loose Source
Routing already present in IPv4. Specifically, it seems like it would
be natural to have a NAT-like box between a local 10.*.*.* network
and the open internet that put the IPv4 address of the NAT-like box in
the source address of the IP datagram and put the local, originating
10.*.*.* address in the reverse route of the Loose Source Routing option.
When this NAT-like box received a return IP datagram, this local
10.*.*.* address would be the next address in the Loose Source Route
specified by the return IP datagram. The NAT-like box would then
forward this IP datagram to the local network, as is specified by the IP
protocol. In this way, you would get an effective address length of 32 + 24.
I don't see why this technique couldn't be applied recursively to get an
unlimited addressing length. Essentially, the Loose Source Route is used
to address entities in dependent address spaces.
I was wondering about this because (and here is my loose link to E) this
address extension mechanism could be bi-directional, unlike the port
mapping NAT protocol that is in common use, where you can only make
outbound connections. If an outside box (say an E Vat) had the complete
address (ie: the Loose Source Route) for a behind-a-NAT box (another E Vat),
then it would be possible for the outside box to initiate a connection
to the behind-a-NAT box. It'd be interesting to know if any existing NAT-boxes
could be duped by Loose Source Routing into providing this behaviour already.
At first blush, it would seem to me that deploying this address extension
mechanism would be a lot easier than deploying IPv6. Does anyone know if
this was ever considered?
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com