Re: [apnic-talk] Layers of NAT
>
> On Nov 15, 2010, at 5:50 PM, Lindqvist Kurt Erik wrote:
>
>> Maybe I am mistaken but I am under the impression that in many networks/countries in Asia, IP connectivity for DSL customers are done using multiple layers of NAT.
>
> The main sectors in which I've see this done is in the wireless mobile space, as well as to some degree within fixed broadband networks within a significant Northern Asia market as well as a couple of smaller Southeast Asian markets.
+1 on mostly seeing it in wireless (cellular, occasionally wimax) networks. Very rarely do you see it in wireline networks, and even then typically only in fairly small stub networks or very isolated and small segments in bigger operators.
>
> Typically, the same RFC1918 space is used over and over again within the CPE/access layer, and then other RFC1918 space is used at the distribution layer. There's also a fair amount of what amounts to internalized route hijacking/misassignment of valid CIDR blocks which are deemed irrelevant or in fact undesirable by some parties, so misusing them for this purpose is seen as inconsequential, if not a positive good in its own right.
The main problem with many RGs in the market that has been raised on other lists is that they often come preconfigured with say 10.0.0.254/8 as the LAN IP address, and don't enjoy being assigned a random 10/8 WAN address via PPP or DHCP. There are some devices where it works acceptably and others where it doesn't, or it behaves erratically.
Operators with those concerns want to minimize that incidence in a field deployment of millions of devices.