Just to be clear when it comes to UDP, STUN doesn't penetrate NATs -- it
merely helps classify them. ICE is what does the actual penetration, but it
doesn't support port prediction, and thus doesn't support
symmetric/symmetric or symmetric/restricted-cone NAT combinations. If
symmetric and restricted-cone make up 13% and 44% of the internet,
respectively, then that means ICE will fail in at least:
Symmetric/symmetric = 13% * 13% = 1.7%
Symmetric/rstr-cone = 13% * 44% = 5.7%
Rstr-cone/symmetric = 44% * 13% = 5.7%
Total = 13.1% of real-world combinations
So an ideal implementation of ICE should get around 87% peer connectivity.
But ideal is really, really hard. I don't know of anyone who's actually
implemented it and tested it on any relevant scale, so I don't know its
actual rate of penetration, but I assume it's far less than that.
-david
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:p2p-hackers-
> [EMAIL PROTECTED] On Behalf Of Stas Khirman
> Sent: Tuesday, January 09, 2007 6:21 PM
> To: 'theory and practice of decentralized computer networks'
> Subject: RE: [p2p-hackers] Re: HTTP design flawed due to
> lackofunderstandingofTCP
>
>
> Alexander wrote:
>
> >Please note also, to the best of my knowledge, there is still no
> >published NAT traversal scheme, comparable on its quality to, say,
> >Hamachi. Published NAT traversal mechanisms, like ICE, lacks some
> >important knowledge, critical for the high-probability NAT traversal.
>
> Am I mistaken, but STUN and STUNT claim to penetrate almost 100/90%
> (respectively) percents of various NAT combinations? Am I missing
> something
> here?
>
> http://nutss.gforge.cis.cornell.edu//pub/imc05-tcpnat.pdf
>
>
>
>
> _______________________________________________
> p2p-hackers mailing list
> [email protected]
> http://lists.zooko.com/mailman/listinfo/p2p-hackers
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
