Re: v6-v4 transition scenarios, take 1

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 28 jul 2008, at 19:19, Pekka Savola wrote:

> AFAIK, p2p protocols, at least downloading, are smart enough to work  
> through a NAT box that doesn't support any ALG, UPNP, port  
> redirection, etc.  I have personal experience with this :-)

The ones that I've used depend heavily on UPnP (or NAT-PMP) or manual  
port mappings. If you can't receive incoming sessions, it still works  
because most other peers can. But you'll see a marked increase in  
download speeds when you enable incoming sessions because you can now  
connect to a larger set of peers.

> So, does this require special casing in NAT64 techniques or are such  
> approaches enough client-server -like?

For the BitTorrent-like stuff you probably don't really have to do  
anything: peers behind NAT64 wouldn't be able to receive incoming  
sessions from IPv4 peers, but they can still have outgoing sessions to  
many IPv4 peers and they should also be able to talk to IPv6 peers in  
both directions. (Although IPv6 firewalls get in the way, but you can  
disable those.)

For SIP ICE should work, especially with endpoint independent mapping  
in the NAT64 box. Other protocols could be upgraded to use ICE as  
well, but for things like BitTorrent this would require significant  
changes to the signaling protocol.

> > I expect once the IPv6 ball gets rolling people will start turning  
> > off IPv4 surprisingly fast.

> Personally I doubt it -- a lot.  Or we have a different things in  
> mind when we talk about "IPv6 ball gets rolling".

> I wouldn't turn off v4 on any service I have association with until  
> at least 90% and hopefully more than 99% of the user base (or  
> potential user base) could reach it.

Right, I was thinking of the client side. There are basically two  
things that I regularly do that I can't do when I run IPv6-only now:  
printing and VoIP. This will only get better over time. (Or I use a  
much smaller set of apps than most people.)

> > Many of the currently used NAT traversal techniques don't work well  
> > with multiple layers of NAT, and with more users behind an IP  
> > address it will be harder for users to open up ports to host  
> > incoming sessions. So I don't think this is necessarily true.

> I guess you're referring to NAT traversal mechanisms such as uPNP,  
> NAT-PMP etc.

Yes.