← Back to team overview

p2psp team mailing list archive

Re: P2PSP, GSoC 2015

 

On Tue, May 19, 2015 at 9:34 PM Max Mertens <max.mail@xxxxxxxxxx> wrote:

> Hi Vicente and Juan Pablo,
>

Hello!


> To start, currently I am trying to setup a virtual network device
> structure (via iptables and Mininet or Virtualbox machines), to simulate
> and experiment with NAT traversal between machines behind different types
> of NATs.
> Besides that, I am reading up on different NAT traversal techniques (ICE,
> STUN, hole punching).
>

Good idea.


>
> Symmetric NATs are little beasts when it comes to P2P software. :-) The
> chownat and pwnat softwares claim to traverse them as well, so this seems
> to be possible.
> Probably I will first try to implement UDP hole punching, and later if it
> is needed for symmetric NATs, ICMP hole punching as well (though this
> requires a process with CAP_NET_RAW permissions to send raw ICMP packets).
>

Yes, symmetric NATs are difficult to deal with.


>
>
> In a previous email, you mentioned that "*In our case, we can suppose
> that each peer knows each other. The symmetric approach should be enough.*
> "
> So just to be sure, a technique like STUN is not needed, as the hosts are
> supposed to know each other's public IP address (and probably the port)?
>

Yes, the splitter works like a STUN server. When a peer wants to join the
team, the splitter sends to the new peer the list of peers of the team. But
notice that this list will be useful, only of the peers are not behind
symmetric NATs, because the public port that the splitter sees of a
symmetric NAT-ed peer will be different of the public port that the
incomming peer will see of the a symmetric NAT-ed peer.


>
> Kind regards,
> Max
>
>
>
Best,
Vi.

Follow ups