>Known, I'm sure. Note that those are 0+17, 256+17, 512+17, etc. I think it's a collation sequence bug -- are you doing something besides SET COLLATE TO MACHINE? Are you running SP4?

You've got it, the problem happens with SET COLLATE TO "GENERAL" but not with SET COLLATE TO "MACHINE". I'm on VFP6 SP3.

It may be known but it does not appear to be documented. The closest thing I can find in the October MSDN Library is an article entitiled "BUG: Incorrect Return of SubSelect in SELECT-SQL Statement" but what it describes is a bit different than this. (BTW didn't the Knowledge Base used to show the Q number in the title of the article? I don't see this any more in the October 2000 edition.)