A new class of probabilistic decoding algorithms suitable for data network protocols is introduced, which intrinsically combines FEC and ARQ techniques.
The main algorithm is all-purpose, straightforward, quite flexible, and has parallel architecture. The tradeoffs between space/time complexity
and performance are evaluated by simulation using importance sampling. A few results for block codes such as BCH (31,15) and quadratic residues code QR (47,23) are presented.