RaptorQ
RaptorQ is the systematic fountain code specified in RFC 6330 for source-block recovery using original source symbols and generated repair symbols.
RaptorQ encodes source blocks independently. Extra symbols for one block cannot repair another block's missing information. Object-wide packet counts therefore cannot establish complete recovery. Decoding needs the object transmission information and the matching block and symbol identifiers. [1]
Its four-byte FEC Payload ID uses an 8-bit source-block number and a 24-bit encoding-symbol identifier. Bytes 01 02 03 04 identify block 1 and symbol 131844, unlike the Raptor layout. This is a field-layout example, not an encoded-packet validation. [1] [2]
RFC 6330's failure bounds use supported extended block sizes K' and independently, uniformly selected identifiers. They are conditional requirements, not a guarantee for every K-symbol set or a measured storage-loss rate. Source blocks may be padded from K to K'; with multiple sub-blocks, a symbol need not be a contiguous slice of the object. [1]
Verified erratum 5548 corrects the transfer-length limit to 942574504275 octets: 56403 x 65535 x 255. The 8-bit block-count field Z cannot represent 256. The width of a block identifier does not establish the number of blocks an object can contain. [3]
Relations
- related
- Raptor code, Fountain code, LT code, Erasure coding, Systematic code
Limitations
- Correct equation solving cannot establish that accepted data are authentic; verification remains a separate requirement. [1]
Citations
- Luby et al. RFC 6330 (2011), sections 3, 4.4, 5.3.1, 5.8 and 6: RaptorQ blocks, identifiers and conditional recovery.
- Luby et al. RFC 5053 (2007), section 3.1: Raptor uses a different FEC Payload ID layout.
- RFC 6330 technical erratum 5548, section 3.3.2: block-count and transfer-length correction, verified 2 June 2020.
Sources & caveats
- RFC requirements and exact field-layout arithmetic are distinguished from codec execution and storage-fleet evidence.