Trying to understand "normal" setup

[Antti Mäkipää]

I wanted to follow up on the conversation in Github issue here: https://github.com/meetecho/janus-gateway/issues/3166#issuecomment-1429552703

Lorenzo said that in 99% of cases nat-1-1 (or STUN/TURN) is not needed and this made me think I might have misunderstood something.

We have Jannus running on a private server in Google Computing Engine. Our WebRTC client receives an SDP from a legacy SIP device and asks Janus NoSIP plugin to process it. Janus of course needs to add its public domain to the Contact (c=) line for the client to be able to stream media through it.

Or does it? What kind of setups are those 99% where Janus can send a local IP and everything works? How does the endpoint reach Janus? I suppose it's not common for Janus and the WebRTC client to be in the same network, is it?

[Lorenzo Miniero]

Janus is in 99% of the cases on a publicly reachable address, not NATted, which is why most of the times you don't need to do anything fancy with addresses: users are almost always NATted, instead, so they definitely need STUN and/or TURN. You running Janus in a private server behind a NAT are an exception, not the rule, so while configuring STUN and/or nat-1-1 does make sense in your scenario, in the 99% of other more common usages people don't, hence the warning.

L.