The NAT traversal used by WebRTC requires in part a STUN (or TURN)
server clients can ask for their public IP address. So far, we've just
hard-coded a particular public stun server[*] into our code.

While webrtc allows web content to specify its own stun or turn server,
I always expected we would need to deploy and maintain our own stun
servers as a default option. I've volunteered to get something set up
for testing.

Because a round trip to this server is part of call setup, it's better
to have geographically distributed servers to reduce latency. I propose
we use Amazon's EC2 virtual machines for this. They're easy to get
started, Mozilla already has a relationship with them for other other
work, and they have data centres in six locations around the globe.

I've volunteered to set something up for testing. STUN is a very simple
protocol, so this should be inexpensive and easy to set up. I've opened
bug 807494[+] to track this. Please contribute any technical ideas there.

I wanted to mention the project here just so people are aware of it, and
to get feedback on the general idea.

 -r

Cheers
Suhas/crypt

Sent from my iPad

On Oct 31, 2012, at 3:35 PM, "Ralph Giles" <gi...@mozilla.com> wrote:

We'll also eventually want to collect stun prefs from the user and any
platform interfaces.