[BEHAVE] NAT64 mapping behaviour w.r.t. scaling

[Simon Perreault <[e-mail]>]

Hello,

The endpoint-independent mapping behaviour is good for NAT traversal, 
but limits the number of sessions to 65535 per IPv4 address in the pool. 
This limit is low, when also considering the fact that sessions may 
still be active for a relatively long time until they time out.

Since we have the option of doing address-dependent filtering, couldn't 
we also at the same time opt for address-dependent mapping? It would let 
us overload IPv4 source ports for different IPv4 destination addresses, 
which would greatly increase the scalability of the NAT in some common 
situations. I think it would have the same NAT traversability as what we 
already have with address-dependent filtering (am I right?).

I understand this would change a lot of text but I think the scalability 
potential warrants this change. I am willing to contribute text if this 
proposal is received positively.

Thanks,
Simon
-- 
DNS64 open-source   --> http://ecdysis.viagenie.ca
STUN/TURN server    --> http://numb.viagenie.ca
vCard 4.0           --> http://www.vcarddav.org