Itzik Ben-Gan looks at ways recent T-SQL improvements:
One of the complexities that is of a specific importance to this section is comparisons that potentially involve NULL comparands, such as ones that you use in filter and join predicates. Most operators that you use in such comparisons, including the equals (=) and different than (<>) operators, use three-valued logic. This means that there are three possible truth values as a result of a predicate that uses such operators: true, false and unknown. When both comparands are non-NULL, such operators return true or false as you would intuitively expect. When any of the comparands is NULL, including when both are NULL, such comparisons return the unknown truth value. Both filter and join predicates consider unknown as a non-match, and sometimes that’s not the behavior that you’re after.
The distinct predicate (IS [NOT] DISTINCT FROM) should simplify a lot of code in the wild.