Alternative discovery and transport protocols

[vasco-santos]

We have been discussing recently what new protocols for discovery and transport we should aim to support in the near future.

The purpose of this issue is to discuss what are the possible protocols, its advantages according to the ones we currently support, offline-first capabilities as well as their compatibility with devices / types of nodes that we have.

We have been discussing some protocols so far, like bluetooth, audio-frequency, among others, and we have several other ones that we never discussed, such as infrared.

cc @momack2 @mgoelzer @raulk @jacobheun @daviddias @Stebalien @jhiesey

[vasco-santos]

One of the protocols that we have been talking recently is Bluetooth, for obvious reasons. It encompasses discovery and transport, it is widely spread in devices and requires low resources. One of the biggest points for it is the new WebBluetooth api libp2p/js-libp2p#261 which would enable web nodes

In this context, I made some initial experiments with WebBluetooth and here follow my initial report. Basically, I Created a simple React app for finding and getting data from my smartphone through Bluetooth. A mobile app (BLE scanner) was used as my smartphone does not have Bluetooth Low energy support.

Concerns

• Bluetooth 4.0 or later
• Needs to be a Bluetooth Low Energy device (installed an app in my smartphone for testing it, it is not supported for a large number of regular devices), more proper to IoT stuff
• Needs a click on application level (API)
• Only supports Passive scanning for now, you can't scan for all devices around you — for good privacy reasons — and you can't just connect to a single device
• Draft spec
• Can't find in the spec / doc / examples a way to exchange arbitrary data, only to access characteristics of the device, such as heart beat, battery, …

Other notes

• https://github.com/noble/noble Node support for exchange arbitrary data (no browser!)
• Web NFC API also exists
  • Report from w3c created 1 month ago though

Conclusion

• We can try to go deeper and try to create a PoC for js-libp2p using web bluetooth. I still need to make sure if they really do not support the exchange of arbitrary data, but I think it will not be stable and available to a large set of platforms in a near future. However, it seems the best approach for browser nodes in spite of not being in a good shape yet

[lidel]

@vasco-santos I see latest WebBluetooth spec supports reading and writing to "Characteristic User Description":

• /samples/web-bluetooth/read-characteristic-value-changed-async-await.html
• /samples/web-bluetooth/write-descriptor-async-await.html

  This sample illustrates the use of the Web Bluetooth API to write to the descriptor "Characteristic User Description" on a nearby Bluetooth Low Energy Device

If it was possible to expose such a writable characteristic value from a browser (which I don't think is possible atm?), then we would have means of passing arbitrary data between two browsers.

[Tolsi]

You might be interested, there's an article about custom audio transport protocol. Source code

[bertrandfalguiere]

I love the idea behind the audio transport! It may be nice to bootstrap a discovery by advertising an address for another, regular, protocol. I don't see many use cases apart from discovery, but I'm sure more creative people will.

Along the same line, we could even imagine a vibrating phone transmitting data to another's accelerometer. I don't know if it's feasible in practice and/or even useful...

[lidel]

A proof-of-concept for WebRTC signaling using sound: https://github.com/ggerganov/wave-share

[lidel]

Chrome 81 introduces the mobile web to NFC with an origin trial:

• https://blog.chromium.org/2020/02/chrome-81-near-field-communications.html
• https://web.dev/nfc