Hosting IPFS node with ngrok

IPFS-NAT-traversal-using-ngrok.md

Hosting IPFS from behind NAT/Firewall using a free reverse proxy (ngrok)

1. Expose localhost's port 4001 to public internet using ngrok: ngrok tcp 4001. Tip: Use -region= flag for lower latency.
2. Note the hostname and port returned by ngrok in the form: tcp://hostname:port -> localhost:4001
3. Open the ipfs config json file ~/.ipfs/config
4. Edit as follows: Addresses.Announce=["/dns4/put-the-hostname-here/tcp/put-the-port-here"]
5. Save the config file
6. ipfs daemon

@avatar-lavventura

By fully qualified multiaddress, I meant addresses that contain the ip, port and peerID, e.g. /ip4/<ip>/tcp/<port>/p2p/peerID.

If your google node is connected to DHT, your laptop node can get the fully-qualified-multiaddress of the google node from the DHT by referring to its peerID. When google node is not connected to the DHT (--routing=none), the laptop node must be given the full address of the google node by some other means - e.g. you hand typed it into ipfs swarm connect ....

Your multiaddress only becomes public when it is published to the DHT. As long as you remain disconnected to the DHT, your google node is undiscoverable by other nodes. For added safety however you can always use the Private Networks feature. In that way, only your laptop node and google node can connect to each other using a common secret key - forming a private network.

The slow transfer speed beats me. AFAIK, its not a problem with IPFS.

its not a problem with IPFS.

Like it is not problem for the IPFS developers? or slow transfer speed problem was always there for IPFS but accepted as it is?

@avatar-lavventura I meant that I suppose the slow speed is not due to ipfs per se. Because I and some of my colleagues who used ipfs didnt experience very slow speeds once the peers were swarm connected.

Probably there was some issues in the Google instance I am using since 100 MB was downloaded in ~10 minutes even peers were swarm connected.