(Late reply to this)

I agree with Alex that libutp is a different layer and thus not directly 
relevant (though I'd probably use or UDT or something like it).

As for the NAT penetration itself, I haven't seen any open-source 
implementation that has really been tested at scale.  The theory is 
really simple, but the practice is very, very hard.  Given that, it's 
probably easier to build it yourself from scratch than adapt an existing 
implementation, because all the hard work -- optimizing it to work in 
the real world -- will be the same either way, and it's always easier to 
optimize your own purpose-built (and less over-engineered) solution than 
to struggle with someone else's design decisions.

And I would unquestionably ignore STUN, TURN, ICE, P2P-SIP, etc.  What a 
disaster.

-david

On 01/20/2011 03:43 PM, Alen Peacock wrote:
> A question for those who've built NAT traversal for their apps in the
> past: If you were starting from scratch today, how would you do it? Is
> there a library you think has everything you need (libjingle, pjnath,
> amicima, socialvpn, etc)? Or would you roll your own, perhaps
> implementing ANTS
> (http://nattest.net.in.tum.de/pubs/globecom09-draft.pdf) instead of
> ICE? What if you need super-high success rates for direct connectivity
> (+95%)?
>
> Would you use libutp (https://github.com/bittorrent/libutp) or
> libjingle's psuedo-tcp classes, or some other, or just use plain UDP
> and manage segmentation/retransmit/reordering yourself?
>
> I know this is a well trodden topic in p2p-hackers, but things change
> quickly, and although most of the above have been mentioned in passing
> in the past year here, I haven't seen a good drag-down, knock-out,
> "Here's the Best Way To Do It" discussion in a while. Any takers?
>
> Alen
> _______________________________________________
> 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

Reply via email to