title
Watcha

Geo p2p(js WebRTC) social network.

Deprecated

 

This project has deprecated. Use Watcha2 instead. It based on IPFS and OrbitDB, I think it fully consistent with the Watcha’s philosophy.

 

It allows you to watch users around in real-time on the map, messaging with them, filter by tags, create a video broadcasts using WebRTC.

 

The key features:

• User can watch what’s going around on the map page. Moves of other users are displayed on the map.

• User can define a custom filter on settings page. It use tags to define which users can be shown on the map.

• Chat with others by clicking on their markers.

• Create and share a video broadcasts with your device’s camera by one button click. There is no a performance and bandwidth issues because the other user watching your broadcast can share it with other.

 

The key technologies:

• P2P connections and peer -to-server are based on the custom PeerJS build. Key feature of my build, that it parse(with my library called ‘[JSONWorkers]’) and handle all messages in the WebWorkers threads to increase a performance, especially on a mobile devices.

• Use the custom library to create a peer-to-peer(WebRTC) connections without the server. How-to-do demoby Chris Ball

• My own serverless technology to create and distribute a video broadcasts from one peer to another(like Periscope).

• The server side based on the oldest nodejs and io.js builds.

• The server side can be easely distibuted to a several threads (with child_process.fork) on a one server or a several servers (by proxying a client's requests and Redis pubsub as a data bus between servers) with my custom developed libraries.

• Redis is used as a primary storage:

  1. Redis installation for a user data (accounts) storage use a several master nodes, replicated by a slave nodes and watched by the Sentinel. It uses the snapshotting on a disk and AOF for the data persistence.

  2. Redis installations for the current geo information also uses the master-slave replication with Sentinel watching on it. All of data is storing in memory because of the performance reasons. Also i use the techniques to store a geo coordinates as a one numeric hash to reduce a memory utilization. This numeric hash have also encoded by cryptographic to unable а fake geo coordinates by the users. Server decode the received coordinates with the private key to it's origin geo hash.

  3. My libray called ‘[RedisPipelineCommander]’ backed by ioredis increase the performance of a db requests by merging a several requests into a pipeline and send this query to the database as a one. It distribute a requests for a several redis nodes by a db 'KEY' the query is deal with. When the results are obtained from ioredis, the library divides them in the order in which they came, handle an errors, and resolves with it a Bluebird Promises, stored in the waiting queue.

  4. Use custom LUA scripts to decrease a number of db calls.

• There is no registration is needed. Watcha uses OAuth 2.0 for it, backed by hello.js on the client and my custom library on the server.

• For map rendering the leafletjs with some plugins is used. The custom library and the Leaflet.MovingMarker are used to render the user markers and handling of it's position changing..

• To obtain and watching(for changing in real-time) for the user coordinates used a custom library backed by Geolocation API.

• UI is based on jquerymobile and a several custom components.

• I've created Ubuntu install script for easy installation.

 

My libraries in this project:

• RedisPipelineCommander - auto sharding and merge into pipelines of a Redis queries

• PeerJS (forked. Origin is here) - Peer.JS with some custom features

• JSONWorkers- use WebWorkers to parse/stringify JSON

And other. Just waiting for. For now i have no time