Unfortunately SPSS has some stupid solutions for sampling and there is no need for PSPP to follow them. It could be better. :-)

In menu Data - Select cases, someone can select Random sample of cases. One option is to set approximate percent of cases to be sampled, other is to set exactly how many cases should be selected from the first n cases.

Let's say you want to select a fixed number of cases, but from all cases (in fact this scenario is very likely). You must first check how many cases is there in dataset and then enter this number manually.

So it would be just great to have an option to select a fixed number of cases from ALL cases.

Another good thing would also be to have different methods of sampling, for instance "pure" random sampling, PPT (probalility proportional to size), two- or more stages sampling or subsamples (first sample n units, then randomly divide that sample into several subsamples of different sizes), etc.