Well, as has been pointed out already the 5th "correct" option doesn't handle update statements at all and even for inserts it handles only single row inserts.

Nevertheless, it's a good question (despite that flaw in the answer) because it's thought-provoking. And I don't imagine I'm the only one who wasn't previously aware of Hugo's connect item and added a vote as a result of this, so with some luck the question has added a few votes that may help MS realise that this is a serious issue.

It raises an interesting side-issue: how bizarre does a "work-around" have to be before it ceases to count as a real work-around. I guess any answer to that is going to be subjective. For myself, I reckon a filtered index is a good work-around, and a trigger is a passable work-around. But I'll go along with an earlier comment about the computed column solution - why do I want to invent some weird imaginary column to do something that the SQL standard says I can do without it? It works, of course, but I don't like it one bit. Neither do I like the indexed view work-around, although of course that too works; but confusing an index (an access mechanism) with a view (a derived relation) seems to me to be a step too far from the relational model.