I think few days back same conversation happened... Unique constraint is different from unique index. While creating unique index we can create with filter and that will allow more than one NULL value but not with unique constraint.
create table #Test (P int identity primary key, A int, B int, unique(A,B))
insert #Test(A,B) values(1,NULL),(2,NULL),(3,NULL),(NULL,1),(NULL,2),(NULL,3)
insert #Test(A,B) values (NULL,NULL)
select * from #Test order by P
drop table #TestTry running that code and you'll see that despite teh unique constraing column A has 4 rows with NULL in it, and column B has 4 rows with null in it. In fact there are 7 rows and in every one of them either A or B or both are null.
Your statement about only one null being permitted is valid only when the unique constraint applies to a single column; it is not valid when it applies to multiple columns.