David, 90% success in connection rate is very impressive. Are you refering to UDP, TCP or both? As you say, fallback such as relay is essential for 100% connectivity. One cannot use hole punching without a good fallback scheme. regards, suresh
--- David Barrett <[EMAIL PROTECTED]> wrote: > > -----Original Message----- > > From: Melinda Shore [mailto:[EMAIL PROTECTED] > > > > ICE makes me want to pound red-hot nails into my eye sockets. That said, > > 90% really isn't nearly good enough for residential deployments. > > Incidentally, 90% is where my stats come out today -- this doesn't include > restricted-cone/symmetric and symmetric/symmetric penetration, which has > been prototyped but not added to the production system yet. The hope is > this will get us much close to the 99% range, but it remains to be seen by > how much. > > Furthermore, I'd wager that for most applications (such as VoIP), even > routing 1 in 10 calls through a central relay isn't going to break the bank. > (Or for mega-discounts, use the full-cone and non-NAT'd clients (>30%) to > act as "free" relays for their peers.) > > Thus while NAT penetration using what I guess we're calling the "hole > punching technique" has been demonstrated to work at 90%+ on a very large > scale (hundreds of thousands of clients tested) we should expect that it's > possible to get even better still. If ICE can show how it'd do better, > that'd be great. But we can't tolerate ICE being any worse than that, and > thus before rushing ahead with standardizing an unproven ICE, we should just > take the time to prove it. > > -david > > _______________________________________________ Sip mailing list https://www1.ietf.org/mailman/listinfo/sip This list is for NEW development of the core SIP Protocol Use [EMAIL PROTECTED] for questions on current sip Use [EMAIL PROTECTED] for new developments on the application of sip
