On Jan 4, 2008, at 10:00 AM, Rich Shepard wrote:
> Maybe we have a difference in semantics that is dependent upon
> the application.
The distinction can be important, as SQL has only partially
implemented 3-valued logic (TRUE/FALSE/UNKNOWN) and treats NULL in
sometimes unexpected ways. NaN could be a useful value to distinguish
between truly unknown quantities (say, that particular machine was not
taking measurements during a particular test) and those where you've
received a value from a machine and it's NaN. But I agree, it does
depend on the application.
Michael Glaesemann
michael.glaesemann@myyearbook.com