The invention relates to the generation of error correction data for available data packets 53 and to the reconstruction of missing data packets based on available data packets and available error correction data. In order to enable an efficient error correction, the data packets 53 are distributed at an encoding end to at least one block of data symbols 54, each data symbol 54 being of equal size (steps 403,404). Then all data symbols 54 of a respective block are used in an encoding operation for forming a respective one of at least one error correction symbol 56 for this block (step 405). At a decoding end, received data packets 83 are distributed again to a block of data symbols 84 and missing data packets are reconstructed based on these data symbols 84 and on error correction symbols 86 associated to the block.