In the last episode (Feb 01), Carsten Gehling said:
> > >The SQL below illustrates what I believe is a bug in MySQL up to
> > >and including 3.23.47. Essentially I need a unique key where one
> > >or more of the component fields of the unique key can be NULL.
> > >What seems to happen is that you can add "duplicate" rows if the
> > >value is NULL.
> > This is how UNIQUE indexes work in MySQL. All values except NULL
> > must be unique. If you use a PRIMARY KEY instead, this won't
> > happen. Of course, the reason it won't happen is that PRIMARY KEY
> > disallows NULL entirely.
> I thought that UNIQUE indexes constituted what is known in the relational
> model as a "candidate key" and, AFAIR, none of the members in a candidate
> key may contain NULL values.
a UNIQUE index must also have the NOT NULL attribute to be a candidate
key. Otherwise it's just an index.