The text now uses the approach where the probability of using the last source is not given, and it is 1 - sum of all the rest (and it is an error if this is negative).

Is everyone satisfied with this?

Has the implementation been changed?