In this resource the author recommends only using Null*s when a column is nullable, else you will suffer from a lot of code to deal with the Null*s (as I found out today the hard way...). My concern though is that if you only use Null*s for nullable columns because Null*s are a pain in the ass, you then get to enjoy "zero" values creeping in when you don't expect it (with a Null* the insert/update would fail as a null would be sent through rather than a empty string, false or 0). Perhaps this just means that you need to test your code properly. What is the recommended practice here?
Use Null* for types defined as nullable in the database if you need to know when they were null or the default zero value doesn't cut it for you. Avoid them whenever that's not the case.