[p2p-hackers] Symmetric port prediction, state of the art

David Barrett dbarrett at quinthar.com
Thu Sep 28 02:20:50 EDT 2006


Hi Michael.  The problem with a TURN server is you need to route all traffic
through a middleman.  This is slow and inefficient, though admittedly very
effective.  I'm looking for a way that is fast, efficient, and also
effective, and symmetric NAT penetration is the holy grail.

-david

> -----Original Message-----
> From: p2p-hackers-bounces at lists.zooko.com [mailto:p2p-hackers-
> bounces at lists.zooko.com] On Behalf Of Michael Liu
> Sent: Wednesday, September 27, 2006 6:51 PM
> To: theory and practice of decentralized computer networks
> Subject: Re: [p2p-hackers] Symmetric port prediction, state of the art
> 
> I'm a little confused about this issue.
> In my opinion,the most efficient way to establish connectivity between two
> clients both behind symmetric NATs are through Turn server, because the
> external NAT'd address  each client uses with the stun server is different
> than the addresses they use with each other, and only Turn server can
> communicate with the client correctly after the client punched the hole on
> the NAT equipments.
> I think port prediction is less efficient because there are so many ports
> needed to be scanned, and sometimes it doesn't work .Some NAT can use a
> pool of several Global IP address as NAT'd address, so it's possible that
> you can't establish the connectivity even you have scanned all the 65535
> ports.
> 
> Regards
> Michael
> 
> 
> 
> 
> ----- Original Message -----
> From: "David Barrett" <dbarrett at quinthar.com>
> To: "'theory and practice of decentralized computer networks'" <p2p-
> hackers at lists.zooko.com>
> Sent: Monday, September 25, 2006 9:11 AM
> Subject: RE: [p2p-hackers] Symmetric port prediction, state of the art
> 
> 
> > > -----Original Message-----
> > > From: Ryan Barrett
> > > Subject: Re: [p2p-hackers] Symmetric port prediction, state of the art
> > >
> > > however, david already acknowledged rendezvous servers. he's
> interested in
> > > situations when you can't or won't use a rendezvous server, for
> reasons
> > > from
> > > lack of resources to robustness to lawyer-proofing.
> > >
> > > in that case, you're forced to fall back to techniques like port
> > > prediction,
> > > using STUN and friends...which is what he was actually asking about.
> >
> > Actually, to clarify, I'm fine with rendezvous servers.  But even with
> all
> > the servers in the world it's still not possible to connect two clients
> who
> > are each behind symmetric NATs.  This is because the external/NAT'd
> address
> > each client uses with the rendezvous server is different than the
> addresses
> > they use with each other.  Thus -- so far as I know -- even with the
> help of
> > a rendezvous server, some type of port prediction still needs to be
> employed
> > and I'm curious for what others have tried.
> >
> > -david
> >
> > _______________________________________________
> > p2p-hackers mailing list
> > p2p-hackers at lists.zooko.com
> > http://lists.zooko.com/mailman/listinfo/p2p-hackers†Ûiÿ÷'™¨¥É¨h¡Ê&



More information about the p2p-hackers mailing list