Essentially it's a SIP phone. SIP is a protocol that is used to initiate sessions. So if I want to talk to you (usually using Voice Over Internet Protocol), I can use the Signal Initiation Protocol to send a message saying that I want to talk to you. You respond by saying that you would like to talk to. SIP then sends some configuration information about how we're going to talk than then hands everything over to another protocol (usually something called RTP).

Now, one of the problems with SIP is that while it is peer to peer, you have no real way of discovering your peers. Enter Distributed Hash Tables (DHT). I don't really know how this works, apart from it's how bittorrent clients discover each other without a centralised tracker. But essentially, I think you ping out to the internet saying "I want to talk to X". Someone near to you sees it and says, "I'm not X, but I can get you one step closer to X. Talk to my friend Y". And so you repeat that process until your discover where X is.

Another problem with SIP lies in the peer to peer nature. Normally we don't put out computers right on the internet. For one thing it's really hard to get IPV4 addresses any more. So nobody can actually talk directly to you. Instead we are behind routers that have an external address on the internet and map requests to an internal address on your local network. This is called Network Address Translation (NAT).

So, when we are trying to discover where our friend is, not only do we need to use DHT to inch our way towards them, but we have to traverse NAT as well (what the person thinks their address is, is not the address we see because of NAT). Not only that, but we usually add firewalls on our routers because we don't want just anybody to send traffic on our local area networks.

ICE, STUN, and TURN are ways for a client to discover their external address on the internet as well as the port you need to use to talk to them (the port is the most difficult part to figure out). Once they do that, when they send a request to you, they can tell you what address and port to reply to. In many cases this won't work (and I'm informed by very reputable people that this is because router manufacturers fail to follow the specs... I believe this). So instead of jumping through hoops to discover the port that you need to talk to, there are some standards in place that allow a client to talk to the router and say "I'm expecting my buddy to talk to me on this port. Would you please open a big gaping hole and let the traffic through. I promise I know what I'm doing." Now, sensible people do not allow their routers to do this kind of crazy thing because malware is really good at sending requests to routers and promising to be good. So basically, if you know what you are doing, and have control of your router, you can poke holes in it and make this work. If not, then there is a good chance that you won't be able to communicate to people. This is why all the VOIP services out there use a centralised server (because if you carry all the traffic with a centralised server, both ends can contact the server and it can act as a mediator... the downside is that the server has to physically move all the traffic to ensure that it works).

Now what they are saying is that they have hacked the SIP protocol to discover the other peer using DHT. This is possible because SIP is the most under-defined protocol in the world, except for perhaps HTTP (on which is is based). So you can add whatever the heck extensions you like, and nobody cares. Then they use their NAT traversal techniques (which will work most of the time if you know what you are doing and have control of your equipment) to get to the other side.

I am imagining that for messaging they are simply using one of the many SIP messaging extensions (or they have made their own ... which is cool, because SIP is a free-for-all, like I said).

Not sure what else. It seems that they are probably using RTP for the VOIP protocol because they don't say. They are using pretty standard libraries for the codecs for audio and video. This stuff works exceptionally well these days. If you get connected and the ports are all aligned so that you have 2 way audio, chances are you will have extremely good audio/video. The codecs they are using are standard for SIP phones, but if you haven't experienced it before, I think you will be surprised by the quality.

Anyway, even though I probably sound quite negative, I'm actually a pretty big SIP fan (I've written an implementation from scratch before). It's just that I'm not sure that the slowness of DHT combined with the witchcraft that is NAT traversal is going to meet the needs of the average person. One day if we all go IPV6, then we can get rid of NAT and open a single, well defined port so that we can forget all this nonsense. But before that happens, I don't think we'll see a p2p application that will challenge the big services.

https://tuleap.ring.cx/plugins/mediawiki/wiki/ring/index.php...