I, for one, really appreciate the lists of data you've published. My problem, though, has frequently been that it takes way too much time to find a database from those types of samples that resembles what I need close enough. For example, let's say I need to simulate something that needs to have 1 million rows of random but constrained dates and amounts across 2 years and 200 accounts. How long would it take someone to find just the right data? Chances are they could never find it quite the way they want it sooooooo, they may have to find two or three databases and write some code to glean the data in a format they wanted.

Why not just write code to make the data to meet the requirements to begin with? It's really not that hard. I know, I know... what about some random names? Do they have to be real names or just variable lengths of characters that we could derive from a GUID? What about addresses? Again, do they have to be real street addresses or can we generate a random number and concatenate with parts of a GUID. Sure, there will be times when the addresses actually have to be valid for test... then maybe one of those databases you posted will come in handy. But for that date problem you cited? It's easier to build the data than to find it in one of the databases you were kind enough to post.