In recent years an increasing amount of our knowledge about finite groups, and especially the sporadic simple groups, has been obtained by computer calculations. This has many advantages over more traditional methods, especially speed and accuracy, and problems can be solved that are out of reach of theoretical methods. But there are also some disadvantages, the most frequently mentioned being problems of checking or ลฝ reproducing results. The accusation of unreliability need not detain us, as . the average published proof ''by hand'' is equally, if not more, unreliable. Much progress has, however, been made in remedying these deficiencies. A properly carried out computational proof can be much more rigorously and thoroughly checked than any proof ''by hand,'' and if it is properly documented then there should be no problem with repeating the calculations and reproducing the results. It has to be admitted, however, that many computational results fall far short of these ideal standards.
The aim of the present paper is to make a small contribution toward improving the reproducibility of computational results on the sporadic simple groups. True reproducibility requires that both data and programs be produced independently. As regards programs, there are several inde-ลฝ . pendent systems capable of performing basic or not so basic calculations with permutations or matrices. As regards data, the situation is much less satisfactory. Where two independent sets of generators for a given group exist on computers, the two sets usually bear no relation to each other, and it is often all but impossible to obtain one from the other. While it is not to be expected that everyone will agree on what are the ''best'' generators for 505