>>Could be, but I think Hilmar's explanation is even more rational. It would be interesting to hear MS explanations about it. I'm almost sure, Hilmar is correct.
>
>As I pointed out in the Message #715374, his explanation is incorrect.

When I saw your reply, Sergey, I didn't see HOW it made Hilmar's proposition "incorrect". Nor do I see anything in your reply that makes the CONCLUSIVE case that it is for performance reasons alone.

We are all guessing here, I think, and I, like Nadya I think, actually place more weight on the "side-effects" theory than on the "performance" theory.
But I also add that UDFs present similar problems (both types) in Select-SQL commands, and they ARE ALLOWED THERE. Of course there is considerable information forewarning such usage there too.