On Wed, 2007-07-18 at 14:38 -0500, Francois Audet wrote:

> The other thing I find funny in here is that the alternative 
> hole punching techique as described on the web site uses a
> centralize server common to BOTH ends to relay the address, while
> ICE does not (i.e., each end can use their own STUN server).
> 
> In other words, the major difference between ICE and the
> alternative-hole-
> punching technique seems to be that alternative-hole-punching
> assumes both ends use the same STUN server, which makes the 
> rest of the exchange simpler (the "magic box" architectural
> concept is always handy).
> 
> Ironically, this seems to me to be LESS usable in a global network
> where you wont't use centralize servers, especially for p2psip.
> 

This may be sufficient, but I don't think its necessary.  Each UA
punches the hole (and keeps it open, as David described) to whatever
STUN server they like, all that's required is that each server have the
same view of each client.  That is somewhat less restrictive.  In the
common case of the open Internet, as long as the STUN servers used by
each respective UA are in the "open", they don't have to be the same.
After that, the "public" or "visible" IP address/port is put in the
Contact headers.

Regarding the 90% coverage, this is what I'm referring to when I say
that it may be preferable to influence the NAT vendors to not do the
things that put them in the 10% cases.  I sort of thought that was what
the NAT guidelines draft was for.

In addition, if we are standardizing a protocol based on the 10% cases
and imposing potentially huge costs on the 90% to do so, then our
approach is incorrect.  We should be designing for the 90% case and
handling the 10% as exceptional behavior, i.e. a solution where you pay
the big cost only in the case where you are actually in the 10%.

FM




_______________________________________________
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

Reply via email to