I'd love to read similar pointers on sampling in the stream: unknown population size, semi continuous sampling. The limits of your inference and also it's strength.