Why WebRTC DataChannel excites me
Download as PPTX, PDF5 likes4,983 views
This document discusses the potential of WebRTC DataChannels to enable peer-to-peer data transfer between browsers. It outlines how DataChannels allow arbitrary binary data transfer, which could enable new applications like real-time multiplayer games, high-frequency messaging, and decentralized file sharing without servers. However, challenges remain around browser compatibility, binary serialization formats, and issues like NAT traversal that WebRTC does not currently solve. The document argues that within a year most browsers will support DataChannels, fundamentally changing the ability for clients to communicate directly and enabling new networking models for the web.
Why WebRTC DataChannel excites me
- 1. OMG DATA CHANNELS AND HOW TO USE THEM ERIC ZHANG 1
- 2. BINARY TYPES IN JAVASCRIPT Yes binary types exists: ArrayBuffer Blob ArrayBufferView: Uint8Array, etc.. 2
- 3. WHY IS THIS COOL Binary support means we can do anything we want with data 3
- 4. BINARY TYPES IN JAVASCRIPT Do these things with them: URL.createObjectURL(blob) XHR2 WebSocket DataChannel 4
- 5. BACK TO DATACHANNEL Video and audio is pretty cool But I’m excited about DataChannel because 5
- 6. BACK TO DATACHANNEL Arbitrary binary data between arbitrary people That is a lot of power No permissions. No flags. 6
- 7. BUT WAIT THERES MORE Based on UDP – essential when transmission delay matters Configurable with SCTP, for TCP like reliability if you want 7
- 8. WHAT I REALIZED DataChannel fundamentally changes the web. Data has always flowed between server and client. DataChannel allows client to client flow 8
- 9. THE POSSIBILITIES File transfer Faster No privacy issues No bandwidth costs 9
- 10. THE POSSIBILITIES Multiplayer games Substantially higher performance because of lower UDP latency + binary data 10
- 11. THE POSSIBILITIES High frequency events and messaging Higher message rates possible with UDP based transport 11
- 12. PRIVACY Imagine a decentralized Facebook Data actually downloaded from your browser while you browse the web. Seamless experience 12
- 13. DATACHANNEL CHALLENGES 13
- 14. BROWSER COMPATIBILITY Google Chrome - March 7th 2013 GOOD GOOD GOOD BAD No SCTP implementation = No binary, reliable Not coming until later in 27 14
- 15. BROWSER COMPATIBILITY Firefox – March 7th 2013 Nightly 22.0a1 - Several spec issues being fixed RIGHT NOW Bug 837919 - WebRTC: RTCPeerConnection constructor: FQDN not yet implemented (only IP- #s). Bug 846110 - PeerConnection ondatachannel callback receives channel in Firefox, but in Chrome an event is passed in Bug 837035 - Hook up DataChannel SDP to implementation (in particular port and options) Aurora 21.01a2 - Flagless support 15
- 16. BROWSER COMPATIBILITY Binary, reliable, and FF/Chrome interop in stable versions at most a few months away 16
- 17. PROBLEM: JSON I now have a binary DataChannel between two browsers. I want to send some binary data. 17
- 18. PROBLEM: JSON JSON is an extremely bad choice for binary data. Base64 uses 37% more bandwidth and is slow to encode/decode. You can never do this in JSON: JSON.stringify({x: 1, y: binary }); 18
- 19. PROBLEM: JSON Our DataChannel supports binary, why use string serialization at all? Binary serialization options: MsgPack, Thrift, Avro, BSON, ProtoBufs, BERT, and many more 19
- 20. PROBLEM: JSON MessagePack: Most JSON like. Very compact. No schema, not designed for typed languages like others. Shitty for Javascript. JS has distinct string / byte array types unlike many other dynamic languages. 20
- 21. PROBLEM: JSON Solution is simple. Add string type to MessagePack. They hate this. https://github.com/msgpack/msgpack/issues/121 https://github.com/msgpack/msgpack/issues/128 Talk to me about this afterwards 21
- 22. PROBLEM: JSON With a binary serialization format you can do this: blob = pack({x: 1, y: binary }); unpack(blob) -> {x:1, y: binary} Caveat: JS makes serializing part of UTF8 to binary hard. JSON is much better. 22
- 23. PROBLEM: NAT TRAVERSAL WebRTC takes care of NAT traversal for you But not if they are behind symmetric NATs ~8% of connections according to Google Talk dev help page https://code.google.com/p/libjingle/wiki/FAQ 23
- 24. PROBLEM: NAT TRAVERSAL You must provide both a TURN server, a proxy for connections when things don’t work right. No bandwidth , latency savings in this case. 24
- 25. PROBLEM: NOTIFYING CLIENTS In order to establish connection, offer/answer and ICE candidates must be exchanged. Solution: You have to keep a WebSocket or XHR stream or similar open. Consider scaling issues. 25
- 27. THIS ISNT LIKE THE IPHONE Whether you like it or not, almost everyone is going to have WebRTC and DataChannels in their browser within a year. 27
- 28. THIS IS CRAZY The model of the web can fundamentally change 28
- 29. THIS IS CRAZY Someone is going to build the P2P file transfer app that is ridiculously easy 29
- 30. THIS IS CRAZY Someone is going to build the networking library for MMOs in the browser 30
- 31. THIS IS CRAZY Someone is going to build the BitTorrent of the web – just go on the damn page and start downloading movies (I mean Ubuntu ISOs). Crazy. 31
- 32. THIS IS CRAZY Someone is going to build the Node.js of the browser. File/network I/O, Sessions, persistent memory in HTML5 storage, database drivers, etc. 32
- 33. THIS IS CRAZY Someone is going to replace Skype, and fragmented VOIP apps , and telephones in general with the one that just works in your browser. 33
- 34. THIS IS CRAZY Someone is going to build the peer-to- peer CDN of the future. http://peerkit.com 34
- 35. THIS IS CRAZY These aren’t novel ideas. These are real problems. People have been talking about them forever. The technology just hasn’t existed until today. 35
- 36. THANKS FOR PAYING ATTENTION Now go home and code. Twitter: @reallyez Github: ericz I’m currently working on PeerJS and PeerKit. If you’re interested in this stuff, I’d love to talk to you. 36