> > > > > As Bruce suggested that message 1 and 4 will be routed
> > > > > through the overlay.
> > > >
> > > > I don't believe that will work; part of what causes TURN to
> > > > be an effective NAT relay is that the TURN client sends
> > > > a packet directly to the TURN server's publicly-routable
> > > > transport address.  The side effect of the TURN client
> > > > doing that is that the TURN client's (evil, nasty) NAT
> > > > opens pinholes to communicate bi-directionally with the
> > > > TURN server's transport address.
> > >
> > > I agree with you. So I think we should also consider 
> > > NAT's filtering
> > > behavior while we choose a peer to be a TURN server.
> > 
> > The TURN client's NAT filtering behavior??
>
> No, the NAT which the TURN server is behind.

Yes, I agree that needs to be tested.  Some sort of
qualification procedure is needed before a TURN server
advertises its transport address to the overlay.

-d

_______________________________________________
P2PSIP mailing list
[email protected]
http://www.ietf.org/mailman/listinfo/p2psip

Reply via email to