Good question.

I managed to get it wrong (thinking the error terminated the batch - as would an "insert value in identity column with identity insert ON" error, for example - so that the answer would be 1); serves me right for answering before breaking my fast, I guess.

There appears to be a bug in the parser which is shown up by this question (not insisting on a comma which thepublished BNF shows as essential). I guess it's a pretty harmless bug, though.

And why oh why does the explanation refer to sql 2000 documentation instead of to documentation for one of the currently supported releases?