There is no advantage whatsoever. It is, however, useful to know how -sample- and -bsample- work in case you want to do something they cannot do. -bsample-, for instance, draws samples of size n=N from datasets of size of N. Above I've shown how to do this for n<n and="" n="">N.
For instance, pretend I had a dataset of size N, a dataset that produced
suggestive but interesting results. By suggestive, I mean that the standard
errors are too big. I now wonder how large a dataset I would need to measure
the result with sufficient accuracy to determine whether the result was more
than suggestive. I can obtain results for larger N under the assumption "H1:
the suggestive result is true" by bootstrapping samples of size n>N. I could
obtain results for n=2*N, n=4*N, etc. This would be useful if I wanted to
propose spending money to obtain a larger sample. </n>