A Peer Discovery module enables libp2p to find peers to connect to. Think of these mechanisms as ways to join the rest of the network, as railing points.
With this system, a libp2p node can both have a set of nodes to always connect on boot (bootstraper nodes), discover nodes through locality (e.g connected in the same LAN) or through serendipity (random walks on a DHT).
These mechanisms save configuration and enable a node to operate without any explicit dials, it will just work. Once new peers are discovered, their known data is stored in the peer's PeerStore.
For this demo, we will connect to IPFS default bootstrapper nodes and so, we will need to support the same set of features those nodes have, that are: TCP, mplex, and NOISE. You can see the complete example at [1.js](./1.js).
Now, once we create and start the node, we can listen for events such as `peer:discovery` and `peer:connect`, these events tell us when we found a peer, independently of the discovery mechanism used and when we actually dialed to that peer.
For this example, we need [`libp2p-pubsub-peer-discovery`](https://github.com/libp2p/js-libp2p-pubsub-peer-discovery/), go ahead and `npm install` it. You also need to spin up a set of [`libp2p-relay-servers`](https://github.com/libp2p/js-libp2p-relay-server). These servers act as relay servers and a peer discovery source.
In the context of this example, we will create and run the `libp2p-relay-server` in the same code snippet. You can find the complete solution at [3.js](./3.js).
We will use the `libp2p-relay-server` as bootstrap nodes for the libp2p nodes, so that they establish a connection with the relay after starting. As a result, after they establish a connection with the relay, the pubsub discovery will kick in and the relay will advertise them.
Taking into account the output, after the relay and both libp2p nodes start, both libp2p nodes will discover the bootstrap node (relay) and connect with it. After establishing a connection with the relay, they will discover each other.
This is really useful when running libp2p in constrained environments like a browser. You can run a set of `libp2p-relay-server` nodes that will be responsible for both relaying websocket connections between browser nodes and for discovering other browser peers.
## 4. Where to find other Peer Discovery Mechanisms
There are plenty more Peer Discovery Mechanisms out there, you can:
- Find one in [libp2p-webrtc-star](https://github.com/libp2p/js-libp2p-webrtc-star). Yes, a transport with discovery capabilities! This happens because WebRTC requires a rendezvous point for peers to exchange [SDP](https://tools.ietf.org/html/rfc4317) offer, which means we have one or more points that can introduce peers to each other. Think of it as MulticastDNS for the Web, as in MulticastDNS only works in LAN.
- Any DHT will offer you a discovery capability. You can simple _random-walk_ the routing tables to find other peers to connect to. For example [libp2p-kad-dht](https://github.com/libp2p/js-libp2p-kad-dht) can be used for peer discovery. An example of how to configure it to enable random walks can be found [here](https://github.com/libp2p/js-libp2p/blob/v0.28.4/doc/CONFIGURATION.md#customizing-dht).