>>>"very likely to be unique" always worries me. My experience is you eventually get in trouble.
>>>
>>
>>True. And on an afterthought - I don't see why we cannot use a pk - valid for the lifetime of the query - using a derived table
>>
>>

>>	select * ;
>>		from ( select recno() as pk, ;
>>				Test.* ;
>>				from Test ;
>>			) X
>>

>
>I know we can not use RECNO() when select uses more than 1 table. I haven't tested it with the derived table, but I think we need to have a real PK, each table should have a PK anyway.

I know each table should have a pk - but in this case, alas, it does not

Recno() with one table to produce a derived table works fine - take my word for it