This is a follow-up of #36
There are still magic numbers (101, 257 and 100) use to determine the maximum size of a sample.
It would be good to investigate if we can get rid of them and improve the logic of the MaxSize calculation