Re: [MMUSIC] ICE candidate pairing and NAT64

[🔓Dan Wing <[e-mail]>]

On 24-Jun-2015 08:35 am, Jonathan Lennox <[e-mail]> wrote:
> 
>> On Jun 23, 2015, at 9:21 PM, 🔓Dan Wing <[e-mail]> wrote:
>> 
>> On 23-Jun-2015 12:28 pm, Jonathan Lennox <[e-mail]> wrote:
>>> 
>>> Apple’s recent announcement that all iOS apps are soon going to be required to support running behind NAT64 has led me to contemplate what this means for ICE.
>>> 
>>> Should ICE endpoints be required and/or recommended to support using RFC 7050 NAT64 prefix discovery to pair local IPv6 candidates which have a NAT64 gateway with remote IPv4 candidates?
>> 
>> Yes.
>> 
>>> What does this do to connectivity check pacing, or to the total number of connectivity checks?
>> 
>> It means an IPv6-only device with a NAT64 has same number of candidates as a dual-stack device.  I would pace and prioritize NAT64 addresses exactly same as IPv4 addresses would be paced and prioritized on a dual-stack host.
> 
> The issue that occurred to me (having thought about this some more) is that if you’re on a 464XLAT network, I don’t think you want to have the application do its own RFC 7050 pairing — instead, you probably want to use the 464XLAT interface to pair with remote IPv4 candidates.
> 
> Unfortunately, I don’t see any easy way for an application to be able to tell the difference between being on a device with 464XLAT (which will appear as an IPv6 local address with NAT64 plus a private address space IPv4 address) vs. a device with NAT64 on one interface and IPv4 on another (e.g. cellular data plus wi-fi).  

I had remembered 464XLAT used a special IPv4 address for its CLAT, but reading RFC6877 now, I am mis-remembering (or it was deemed a bad idea and pulled from the eventual RFC).

> In the latter case, essentially what an ICE stack needs to do is application-layer 464XLAT for the NAT64 candidates, but doing it redundantly for the former case seems like a bad idea.  Any thoughts, other than deep knowledge of device configuration?
> 
>>> And how many operating systems can support a userspace app doing an RFC 7050 DNS lookup on a per-interface basis?  (Even if you roll your own DNS resolver, you still need to be able to obtain per-interface DNS server configuration information to find the interface’s appropriate DNS64 server.)
>> 
>> Dunno.  MIF was going to solve that.  I will note that dnsmasq appears capable of sending DNS requests out certain interfaces (see the --server option at http://www.thekelleys.org.uk/dnsmasq/docs/dnsmasq-man.html), so it must be relatively possible?
> 
> It’s not just a matter of which interface you use to send the request, but which interface’s configuration information you use to discover the DNS server.  (Consider a dual-SIM device which can connect to  two different carriers’ cellular data networks, both of which are using NAT64.)

Yes, good point.  I agree if all the DNS & DNS64 servers learned from each interface is munged together (like it is with the traditional /etc/resolv.conf), we can't be successful at building synthetic NAT64 addresses.  That's what MIF was going to fix.  I know Dmitry Anipko was making good progress at keeping the learned DNS server from each interface separate on Windows, mostly for split-horizon support [e.g., enterprise VPN] but per-interface DNS64 is exactly the same split-horizon problem.  The learned DNS servers, and the interface each DNS server was learned from, need to be available to the application.  On OSs where that isn't possible, I think worst case is some additional ICE candidates and additional ICE connectivity checks, which will fail?

-d