An alternative full node bitcoin implementation in F#. The source code is on GitHub and this documentation was generated directly from the code.
This project is under development and currently in alpha stage. Generally speaking, writing a fully compliant bitcoin node is extremely hard - some think impossible. The Bitcoin project is experimental and never aimed to deliver a 100% accurate specification. The Satoshi client is the de-facto standard. By far the most used client on the network, it dictates what should be accepted and what should be rejected. Even the slightest deviation can cause a fork and potential loss of funds. It should go without saying
Do not use this code in production when real funds are at stake.
A lot of effort has been put into aligning its behavior with the core client but nevertheless there are almost certainly bugs that could be exploited to cause it to deviate from the reference implementation.
That being said this code has been through the "official" block acceptance test and has run for an extended period of time. It implements all the consensus rules including the bugs that are now included in the blockchain since the genesis of Bitcoin.
However, Bitcoin F# has fully validated the existing mainnet blockchain and passes all the integration tests including large reorg tests. It is also the only implementation in a functional language and comes under 2.5 kLOC, making it the smallest client too.
In writing this code, my goal was
- to make a client that is compatible with the rest of the network to the extend of the supported features
- Simple and readable code. It is not code golf but it is not paid by the line code either.
- Header first, blocks later. The client downloads the block headers first and checks the validity of the chain then it downloads the blocks from multiple peers in parallel. It dramatically improves synchronization.
- Easy pruning. Blocks are stored on disk in a normal directory fashion and can be removed by deleting files. Other implementations use a database or custom data files.
- I didn't want any wallet spending functionality even if I have included support for BIP32, Armory and Electrum style watch-only wallets.
- No JSON API either. I will prefer to extend the core.
All in all, it is a small but working client that can be used as a relay for transactions since it validates inputs and as the base of an online wallet. However, at this point the only wallet features are watch-only for BIP-32, Electrum and Armory wallets.
Further documentation can be obtained at the project github pages
License MIT