On Friday, June 26, 2015 at 10:05:29 AM UTC-6, Ben Schwartz wrote:
UPnP is indeed problematic, but the IETF has proposed a modern standard mechanism for the kind of NAT control that Andrew is interested in:
NAT-PCP. That might be a good place to start.
There's no way we can guarantee that a particular patch will be accepted, but I think some research into NAT-PCP for WebRTC would be welcome.
PCP is just a slightly more recent version of Bonjour, and suffers from a lot of the same design flaws as UPnP does (clear inversion of security, poor auth model, numerous avenues for abuse, and extremely inconsistent router implementations). Also I'd think WebRTC would want to avoid any more potential security threats (be they legitimate or sort of silly, like the public IP address leak the internet was in a huff about earlier); every time that happens some number of users literally disable UPnP in Chrome and FireFox.
But the biggest reason these technologies aren't compelling is ICE is already pretty effective in SOHO environments. It's least effective in larger enterprise settings, but neither UPnP nor PCP are even remotely effective in those environments. Is there some evidence this technology would improve P2P connection rates? Because that's kinda where the rubber hits the road...
Of course research is always a good thing, but honestly I don't think either of these technologies are compelling. My $.02, having worked with both of them in the past.