The authors show that for linear block codes defined over extensions of GF(2) a variant of the sub-optimal soft-decision Dorsch (1974) algorithm offers very good performance with low complexity. Furthermore, we show that, owing to the nature of the algorithm, it can easily be adapted to produce a soft-decision output. This is then exploited in an iterative decoding scheme for product codes which is based on a decoding algorithm first proposed by Pyndiah (see IEEE Trans. Commun., vol.46, no.8, p.1003-10, 1998 and by Aitsab and Pyndiah (see Proceedings of the IEEE Globecom'96 conference, vol.1/3 p.121-5, 1996). Two different interleaver structures are presented yielding different performances both in terms of coding delay and bit-error-rate for a given signal-to-noise ratio. Reed Solomon codes are used in the simulations. For reasons of complexity, only codes defined over GF(16) are considered. Simulations were carried out for the AWGN channel.