[arin-ppml] ARIN-prop-127: Shared Transition Space for IPv4 Address Extension

[Lee Howard]

> > It's ironic that NAT  traversal hacks can't handle NAT.
> Not really. They originally  handled it pretty well, with ALG support 
> making it a lot more friendly.

ALG isn't NAT.  And without defining a layer-violating protocol,
(an application-agnostic application-layer gateway)
it isn't realistic for every NAT implementer to build an ALG for 
every application that needs one.  

> However,  CPE's started utilizing uPNP 
> and applications decided that it was saturated  enough in the market 
> to use it instead of relying on ALG support. This lead to  applications 
> creating non-uPNP unfriendly NAT hacks. Now they will regret  it.

Even if every NAT vendor built an ALG for every new app, new 
apps couldn't be deployed because of the installed base of NATs
which didn't have the ALG for the new app.

So here we are:
uPNP doesn't work for large-scale NAT, because it can't traverse
  the NAT layers 
ALGs aren't a solution, because there are too many applications
  needing gateways
PCP is too late for implementation in applications and appliances

Seriously, for things that don't work beautifully through multi-layer
NAT, IPv6 is the only way to go.  And if you think NAT444 is
your solution for exhaustion, you really need to do IPv6, too.

Lee