I think the point was to simply describe a safer, simpler construct for these common scenarios. The tables were not intended to be exact representations of best practices, but instead simple to understand and use for the examples. Having worked with several vendor-supplied databases over the years, I run across this scenario (two-table filtering with a null-containing column) frequently. I WISH every SQL DB out there was built "properly" or that I had the time and permission to correct shortcomings, but that's far from reality, IMO. Articles like this help move us towards that ideal while offering a way to help protect us from the current reality.