• edwardwill - Friday, July 6, 2018 2:19 AM

    When designing a table, it is essential to realise that there are real-world circumstances where NULL may be an important and meaningful representation of the data being modelled.

    Exactly, they're not evil or anything to me - the important point for me is - is this the best model of the circumstances we can make? Yes, we may need to allow for them or do something funky when aggregating but if that is the correct model, it's the correct model. In general getting the model correct makes design issues sort themselves out most readily.