BitTorrent's DHT is probably one of the most resilient and censorship-resistant networks on the internet. PeerTweet uses this network to allow users to broadcast tweets to anyone who is listening. When you start PeerTweet, it generates a hash
@33cwte8iwWn7uhtj9MKCs4q5Ax7Bwhich is similar to your Twitter username (ex.
@lmatteis). The difference is that you have entire control over what can be posted because only you own the private key associated with such address. Furthermore, thanks to the DHT, what you post cannot be stopped by any government or institution.
Once you find other PeerTweet addresses you trust (and are not spam), you can follow them. This configures your client to store this user's tweets and broadcasts them to the DHT every once in a while to keep their feed alive. This cooperation of following accounts, allows for feeds to stay alive in the DHT network. The PeerTweet protocol also publishes your actions such as
I just followed @919c..or
I just liked @9139..and
I just retweeted @5789... This allows the possibility for new users to find other addresses they can trust; if I trust the user
@6749..and they're following
@9801.., then perhaps I can mark
@9801..as not spam. This idea of publicly tweeting about your actions also allows for powerful future crawling analysis of this social graph.
How does it work?
PeerTweet follows most of the implementation guidelines provided by the DHT RSS feed proposal http://libtorrent.org/dht_rss.html . We implemented it on top of the current BEP44 proposal which provides
put()functionality over the DHT network. This means that, rather than only using the DHT to announce which torrents one is currently downloading, we can use it to also put and get small amounts of data (roughly 1000 bytes).
PeerTweet differentiates between two types of items:
- Your feed head . Which is the only mutable item of your feed, and is what your followers use to download your items and find updates. Your head's hash is what your followers use to know about updates - it's your identity and can be used to let others know about your feed (similar to your
@lmatteshandle). The feed head is roughly structured as follows:
"d": <unsigned int of minutes passed from epoch until head was modified>,
"next": <up to 80 bytes of the next 4 items in the feed, directly 1,2,3 and 4 hops away. 20 bytes each.>,
"n": <utf8 name of the feed>,
"a": <utf8 http url of an image to render as your avatar>,
"i": <utf8 description of your feed>
- Your feed items . These are immutable items which contain your actual tweets and are structured:
"d": <unsigned int of minutes passed from epoch until when item was created>,
"next": <up to 80 bytes of the next 4 items in the feed, 1, 2, 4 and 8 hops away. 20 bytes each.>,
"t": <utf8 contents of your tweet. up to 140 chars>
The reason items have multiple pointers to other items in the list is to allow for parallel lookups. Our skip list implementation differs from regular implementations and is targeted for network lookups, where each item contains 4 pointers so that when we receive an item, we can issue 4
get()requests in parallel to other items in the list. This is crucial for accessing user's feeds in a timely manner because DHT lookups have unpredictable response times.
When you follow someone, you're essentially informing your client to download their feed and republish it every so often. The DHT network is not a persistent one, and items quickly drop out of the network after roughly 30 minutes. In order to keep things alive, having many followers is crucial for the uptime of your feed. Otherwise you can still have a server somewhere running 24/7 which keeps your feed alive by republishing items every 30 minutes.
$ npm install
Installing native modules
The app comes with some native bindings. I used this code to make it run on my computer:
npm install --save-dev electron-rebuild
# Every time you run "npm install", run this
# On Windows if you have trouble, try:
ed25519-supercopto work on Windows I also had to install
node-gypand all the Python2.7 and Visual Studio stuff which node-gyp requires: https://github.com/nodejs/node-gyp
Then run these commands to build it on Windows:
npm install -g node-gyp
HOME=~/.electron-gyp node-gyp rebuild --target=0.36.9 --arch=x64 --dist-url=https://atom.io/download/atom-shell
Run this two commands simultaneously in different console tabs.
Note: requires a node version >= 4 and an npm version >= 2.
$ npm run hot-server
$ npm run start-hot
Toggle Chrome DevTools
- OS X: Cmd Alt I or F12
- Linux: Ctrl Shift I or F12
- Windows: Ctrl Shift I or F12
Toggle Redux DevTools
- All platforms: Ctrl+H
If you use any 3rd party libraries which can't be built with webpack, you must list them in your
You can find those lines in the file.
// put your node 3rd party libraries which can't be built with webpack here (mysql, mongodb, and so on..)
CSS Modules support
Import css file as css-modules using
To package apps for all platforms:
$ npm run package
$ npm run package-all
- --name, -n: Application name (default: ElectronReact)
- --version, -v: Electron version (default: latest version)
- --asar, -a: asar support (default: false)
- --icon, -i: Application icon
- --all: pack for all platforms
electron-packagerto pack your app with
--alloptions for darwin (osx), linux and win32 (windows) platform. After build, you will find them in
releasefolder. Otherwise, you will only find one for your os.
releasefolder and devDependencies in
package.jsonwill be ignored by default.
Default Ignore modules
We add some module's
peerDependenciesto ignore option as default for application size reduction.
babel-coreis required by
babel-loaderand its size is ~19 MB
node-libs-browseris required by
webpackand its size is ~3MB.
Note: If you want to use any above modules in runtime, for example:
require('babel/register'), you should move them form