If a host has several interfaces, each interface must have its own
private IPv4 address (case b) at this interface), or its own IPv4
address, possibly with a restricted port set (case a)).
The host must handle ports according to what applies to its chosen
interface.
Agreed. But it's not "the host", rather it is the application running
on the host that needs to have that awareness. Said another way, it
cannot mindlessly bind to port 0, like applications do today -- but
has to bind to a free port on a specific interface.
This isn't AFAIK specific of 4rd scenarios.
It is a problem common to all of the port-restricted scenarios if we move
the port restriction into the host (A+P, 4rd, dIVI). Either the application
is aware of the port restriction (which is how you avoid an ALG in the host)
and require the application to change, or the application is unaware of the
port restriction (which means an ALG is necessary, for some applications, in
the host) but the application doesn't have to change.
So -- which way is the architecture going to go? I have not seen this
written down anywhere.
-d
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf