HTorrent

HTorrent is a work-in-progress BitTorrent client written in Haskell. This was built as a learning exercise. It is not intended for use in production.

Installation:

NOTE: HTorrent uses pwrite & pread under the hood to allow for parallel reads of tracker content. For this reason, only POSIX system are supported (sorry windows).

Binary Install (Mac)

1. Run wget https://github.com/nicksanford/htorrent/raw/master/release_builds/osx/htorrent && chmod +x ./htorrent
2. Run htorrent --help for usage instructions

Binary Install (Linux)

1. Run wget https://github.com/nicksanford/htorrent/raw/master/release_builds/linux/htorrent && chmod +x ./htorrent
2. Run htorrent --help for usage instructions

Stack install

1. Install / make sure you have git and Stack and that binaries installed via stack install are reachable in your path.
2. Run git clone https://github.com/nicksanford/htorrent && cd htorrent
3. Run make test, all tests should pass. If they do not, please open an issue.
4. Run make install
5. Run htorrent --help for usage instructions

Goals:

I built this as an exercise to learn Haskell (which had only minimal exposure to prior to this project) and use it to build a non trivial project from scratch. I also wanted to learn about parsing, bit manipulation, socket programming, property based testing, how to build efficient concurrent / parallel programs in Haskell, and how to profile & optimize Haskell programs.

Features

Implemented:

• Able to resume partial download with no duplicate work
• Accepting incoming connections from peers
• Computes & sends initial bitfield message to peers
• Hash checking to verify that each piece received is valid
• Is able to seed & leach from itself
• Never downloads the same piece twice.
• Only requests a piece from a peer if it already knows the peer has the piece.
• Parallel leaching & seeding
• Property & unit tests for parsing logic
• Reading single torrent
• Tracks if peer is choking and does not request pieces if peer is choking
• Tracks which pieces each peer has by parsing have & bitfield messages.
• Support for large single files
• Signal handling for Ctrl-C & clean shutdown
• Refactor Parser module to use attoparsec as profiling discovered explicit fold / unfold parser to be the current bottleneck

TODO (next to be implemented):

• Publishing have messages to all peers as pieces complete
• Provide visualization (presentation mode)
• Fix issue where if the tracker server doesn't exist, it does not just blow up. Report back HTTP Call 'http call' to tracker server 'tracker server url'
• Add CI/CD

Not (yet?) Implemented:

• Asking for rarest blocks first.
• Close connections with peers when you have everything and they have everything
• Convert cabal file into an hpack file
• DHT support
• Endgame - asking for the same pieces from multiple peers when number of outstanding pieces is lower than number of peers which can serve them.
• Game theory algorithm for deciding when to choke peers (currently peers are always unchoked)
• Only sending interested messages when the peer actually has pieces HTorrent needs (currently HTorrent is always interested in all peers)
• Property / integration tests for bittorrent FSM itself
• Property tests for RPC parser
• Provide ability to get debug logging
• Provide option to exit on completion
• Requesting tracker for more peers when HTorrent runs low on peers
• Resolve all compiler warnings
• Support for WebSeeds http://techie-buzz.com/how-to/what-are-web-seeds-bittorrent.html
• Support for multiple torrents
• Support for multiple trackers
• Support for specifying the download directory (currently it just downloads the file in the current directory)
• Torrents which contain more than one file
• Track download statistics to report back to tracker
• Tracking & reporting total download & upload rate
• UDP Tracker Protocol

Contributions:

• PRs welcome! If you would like to add/suggest a feature, please feel free to submit a PR/issue.