Oh yea, something worth mentioning is that my analysis did not consider what memory was shared or not. That might influence the actual numbers. I believe the amount of memory used by the weatherd daemon also includes code and not just the heap memory, so the libraries that weatherd loads would be included in that.