First Normal Form

Because the relational database (as least logically) links data only through data values, there must be
something in the data that lets you find a specific row when you want it. This identifying key may be one
column (perhaps ORDER_ID in an ORDERS table), or it may require multiple columns (such as the combination
of ORDER_ID and ORDER_LINE_NUMBER within an ORDER_LINES table.)
We’ll be using this concept of keys throughout the discussion of normalization in the following sections.
There are several forms of keys (primary keys, candidate keys, foreign keys) that we’ll address more pre-
cisely when we get to the physical design stage. For now, we’ll be using the general term” key” to
loosely refer to the primary key of a physical table.

So, for first normal form, each column must be related to the key of the table, but more than just related—
each should be dependent upon the key for its identity. Another way of stating this is that for a table in
first normal form, if you know the value of a row’s key, then you can determine, without ambiguity, the
value for each column in the table. Once you find a row in that table using a key value you have also
found the single value for each of the other columns of that row.