asymmetric NAT

[Eugene Grosbein]

"Bjoern A. Zeeb" wrote:

> > Let's consider a simple scheme with two NAT boxes
> > where packet flow is asymmetric:
> >
> >      A----+
> >      |    |
> > S ---+    T
> >      |    |
> >      B----+
> ...
> > A has 2.2.2.2 for its outer interface, B has 3.3.3.3 for its.
> > A and B both do "static NAT" for S, they translate
> > 192.168.1.1 to 4.4.4.4 (and vise versa). One can try
> ...
> > AFAIK, libalias and ipnat do not support this configuration currently.
> > I'm trying to patch libalias to support this and have some progress
> > but still cannot make work active mode FTP transfers when S is a client
> > and T is a server.
> >
> > Should this schema work in a theory at least?
> 
> the only thing I can think of is to have some kind of protocoll
> beteween A and B that
> 
> a) in almost realtime syncs states
> or
> b) queries the other for a known state about the connection in
>    question and updates it's internal "tables".
> 
> both are problematic and normally addressed in HA software.
> 
> For you scenario an unidirectional syncing would be enough but
> if you want to dtrt do it bidirectional because you might not be able
> to garantee 100% that all traffic leaves through A and responses
> always come in via B.

You are right, packet flow can change.
But why may I need to sync states of NAT boxes in case of static NAT?

Eugene