Mark T.B. Carroll wrote:
> My ISP assigns me a public static IP address but for my gateway
> machine's WAN connection they give me a static RFC1918 address. They
> must do some static transformation of my packets at their end, mapping
> between the RFC1918 address and the public address.
So they are doing NAT from the RFC 1918 address to the public address.
> This messes up my ability to use connection tracking to support
> protocols that cross my gateway's IP masquerading: it's putting the
> wrong address into the protocols when it rewrites the content.
Why?
If your traffic is outbound, you put the WAN address into the
outbound packets and into the great world they go. In the
same way, the reply packets get translated by the ISP and you
can re-translate from the WAN RFC 1918 address into your own
internal addresses.
There's a limitation: Your internal network must not use
a RFC 1918 address conflicting with the ISP's address.
The system limits your inbound traffic, so it may be more
difficult to have a server at your connection - and precisely
this may be the base cause of the address translation at
the ISP.
If the ISP's address translation is pretty constant and one-to-one,
a dynamic DNS solution might work (e.g.
http://www.dyndns.org/).
> Is it possible to have it write the public IP address into them somehow?
> I can imagine I can do it if I give its WAN interface the public
> address, then have another not-connection-tracking 'outer gateway'
> between the gateway and my ISP that does the reverse of the
> transformation they do.
>
> But, can I achieve the same effect without needing another machine?
Not that way.
> (Or another ISP. (-
A suitable bunch of money might make the miracle
with your current ISP also.
--
Tauno Voipio
tauno voipio (at) iki fi