RDBMS Codd rules

Kevin is right!!

Rule 1 - ONLY place 1 thing in each field (first name, area code, color, size, or WHATEVER!) Your data MUST be "atomic", or elementary so it can't be broken down any further.

Rule 2 - ONLY related data is stored in any Table. Every COLUMN in a Table must contain a single atomic Fact related to the Tables "subject" (i.e., Employees Table = LastName, ShoeSize, Age, DateOfBirth, etc.)

Rule 3 - Every record in a Table MUST represent a UNIQUE instance of the Table's subject AND each recors (or row) MUST contain at least one unique index value so the database knows which row you want.

Data Modeling and/or database Normalization have NOTHING to do with any database product
and everything to do with database engine efficiency. Don Quigg

Kevin is exactly right, Boyce Codd normalization is determined by design not by the RDMS. Databyte has some point correct but has also missed some. The distinguishing point between Boyce Codd and 3rd Normal form is that each and every attribute ofa table must be completely functionally dependent upon the primary key. Also the primary key should be a minimal one, meaning don't stick two field together and call them a key just so that every other attribute in the table will be dependent upon the primary key.
Boyce Codd has its place but very few databases will be that highly normalized. Keep in mind that to acheive Boyce Codd that even in a table that had addresses you would not store city and state if zip code were to be stored there. Why? Because city and state can be determined by zip code and they would therefore not be fully functionally dependent upon the primary key. Zip_code would have to serve as a primary key in it's own world.