Jim,

Thanks for the clarifications. I have trouble seeing things in terms of the definitions - your example certainly helps. I have my own set of rules which I think pretty closely match the first 3 forms:

1) Don't repeat fields (i.e. category1, category2, etc.).
2) Break down all tables so that each row has a primary key which represents a quanta of data (i.e. for someone's name).

Question: In your example, does having the title/revision combination violate the 4th normal form, since the primary key consists of a combination of fields? Based on my simple set of rules, it would be kosher, since it's a quanta of data. I guess I don't really follow all the terminology of the definition.

Also, I don't think I get the 5th normal form. Any ideas on alternative ways to think about it?


>David,
>
>Ok here's a brief intro to normalization.
>
>

>1st Normal Form: There are no repeating or multivalued fields in the
>                 table's structure.
>
>2nd Normal Form: All fields are dependant on the entire primary key for their
>                 values.
>
>3rd Normal Form: All fields are dependant on ONLY the primary key for their
>                 values.
>
>Boyce-Codd Normal Form: The table is in 3rd normal form for all feasible
>                        candidates for the primary key.
>
>4th Normal Form: There are not independantly multivalued components of the
>                 primary key (for tables where the primary key is comprised
>                 of two or more component fields).
>
>5th Normal Form: There are no cyclical binary dependencies within the primary
>                 key (for tables where the primary key is comprised of three
>                 or more component fields).
>

>
>So from this you can see that the structure you described is not wholely a result of normalization. Example: A tabel of books where you have fields like this;
>
>Field
>BookTitle (this with Revision is the PK)
>Revision (this with booktitle is the PK)
>Author1
>Author2
>Author3
>Category1
>Category2
>Category3
>DatePublished
>Copyrightdate
>PublisherID
>PublisherName
>PublisherAddress
>
>Now the field Author1 - 3 and Category1 - 3 are multivalued attributes and they violate 1st normal form.
>
>The field CopyrightDate is only dependant on the booktitle alone and not the combination of the title and the revision therefore it is in violation of 2nd Normal Form.
>
>The publisher's name and address are codependant on the publisherID, which is not the PK for this table, and therefore is in violation of 3rd normal form.
>
>The see Boyce-Codd, 4th and 4th NF's we would need to invent different table structures.