A transform stream module for node.js.

Node is all about I/O, so this might help someone who was trying to do I/O over a crappy channel. I don't have any specific use for it at the moment.

npm install fec-stream

var fec = require('fec-stream');

// For Hamming(8,4)
var encoder = new fec.Encoder({method: fec.methods.hamming84});
var decoder = new fec.Decoder({method: fec.methods.hamming84});

// For Reed-Solomon
var rsEncoder = new fec.Encoder({
  method: fec.methods.reedsolomon,
  ecclen: 32 // Number of error correction bytes (default: 32)
});
var rsDecoder = new fec.Decoder({
  method: fec.methods.reedsolomon,
  ecclen: 32 // Must match the encoder
});

// For RaptorQ
var raptorEncoder = new fec.Encoder({
  method: fec.methods.raptor,
  overhead: 0.2, // Redundancy factor (e.g., 0.2 for 20% overhead)
  symbol_size: 1400 // Size of each symbol in bytes
});
var raptorDecoder = new fec.Decoder({
  method: fec.methods.raptor
});

readableStream.pipe(encoder).pipe(decoder).pipe(writableStream);

The available methods at this time are:

• Hamming(8,4)
• Reed-Solomon
• RaptorQ

This guide explains how to add new Forward Error Correction (FEC) methods to the fec-stream library.

The library is designed to be easily extensible with new FEC algorithms. The core logic resides in lib/fec.js, which manages the stream-based processing and dispatches the encoding and decoding tasks to the appropriate method.

Each FEC method is implemented as a separate module in the lib/methods directory. This modular approach keeps the concerns separated and makes it easy to add new methods without modifying the core library.

To add a new method, you need to create a new file in lib/methods that exports an object with the following functions:

• initEncoder(options): (Optional) A function to initialize the encoder. It is called in the Encoder constructor and receives the options object. You can use this to set up any necessary state on the stream instance (this).
• encode(data, encoding, done): The main encoding function. It is called as the _transform method of the Encoder stream.
• flushEncoder(done): (Optional) A function to handle any remaining data in the buffer when the stream is flushed. This is particularly important for block-based algorithms.
• initDecoder(options): (Optional) A function to initialize the decoder. It is called in the Decoder constructor.
• decode(data, encoding, done): The main decoding function. It is called as the _transform method of the Decoder stream.
• flushDecoder(done): (Optional) A function to handle any remaining data in the buffer when the stream is flushed.

The Reed-Solomon implementation in lib/methods/reedsolomon.js serves as a good example of a block-based algorithm. It uses a buffer to accumulate data until it has a full block, then encodes or decodes it. The flushEncoder and flushDecoder functions are essential for handling the final, partially-filled block.

Once you have implemented your method, you need to integrate it into the main library:

1. Add to methods object: In lib/fec.js, add your new method to the methods object. The key should be a short, programmatic name, and the value should be a human-readable name.
2. Require the module: In lib/fec.js, require your new method module and add it to the transforms object, using the same key as in the methods object.

The library uses mocha for testing. When adding a new method, you should also add a corresponding test file in the test directory. The tests should cover:

• Encoding and decoding of small and large files.
• Error correction capabilities.
• Edge cases, such as files with sizes that are not multiples of the block size.

The test suite includes a fun and practical test using the text of the January 1900 issue of "Popular Science Monthly" to test the handling of larger files.

Run tests with mocha.

from http://science.nasa.gov/media/medialibrary/2011/04/13/NASA_Comm_Services_EV-2_TAGGED.pdf:

Most missions employ error detecting – error correcting codes to substantially improve telemetry link performance. DSN users are reminded that their encoders should conform to the CCSDS Telemetry Channel Coding Blue Book (CCSDS 231.0-B-1, September 2003. Acceptable codes include: 1) Convolutional r = 1/2, k = 7 only; 2) Reed-Solomon 223/255 only; 3) concatenated Convolutional / Reed-Solomon and 4) Turbo codes with rates: 1/2, 1/3, 1/4, or 1/6, block sizes: 1784, 3568, 7136, and 8920.