>Ahh, it doesn't have to damage routing transparency. If we were to use >a signaling protocol that is carefully crafted to preserve routing >transparency (e.g. RSVP) then we can avoid this issue. That's what I'm working on, but midcom and upnp as they're currently defined most certainly do have routing-related problems. >The upnp guys are not really thinking of damaging routing transparency. Of course they weren't. But the assumptions that the network is single-homed and that there's only one NAT in the path and that there are no firewall interactions are inherently non- general, and any assumption that they fix "the" problem is necessarily incorrect. Seeing this stuff touted as a general- purpose fix makes me very uncomfortable. Melinda